Can What Has Been Learned Be Let Go? Evidence on Unlearning from Research on Memory, Conceptual Change, Habits and Skills, and Correction

What does “forget first, then learn again” actually involve?

When people move to a new way of working, they are sometimes told to discard the old way first and then learn again. The term unlearning has been widely used in management and human resource development to refer to this discarding side. However, the term alone does not specify what discarding involves. It may mean that what was learned disappears from memory, that it is no longer retrieved, or that it loses out to another idea.

The learning sciences and cognitive psychology contain several research streams that have examined these distinctions experimentally. Memory research has examined how retrieval makes other memories harder to recall, how trying not to retrieve produces forgetting, and how responses that were supposedly eliminated return. Conceptual change research has examined what happens to naive ideas in people who have studied science. Research on habits and skills has examined the difficulty of replacing an established procedure with another. Research on correcting misinformation has examined why people who receive a correction continue to rely on the old information. This note reads these streams with a single question. In the learning sciences, how has letting go of or replacing what was learned been defined, by what mechanisms does it occur (or fail to occur), and what helps?

An essay that questions the assumption that expertise develops in one direction only was written in Is Expertise a One-Way Street? Unlearning and the Blind Spot of Reversible Mastery. This note does not propose a hypothesis; it reports what the literature measured and what it showed.

Scope and method

• Target: peer-reviewed articles, meta-analyses, systematic reviews, and reviews that measured whether, and under what conditions, old knowledge, ideas, procedures, or habits were replaced by new ones in individual learning. Four streams were covered: memory (retrieval-induced forgetting, retrieval suppression, extinction, reconsolidation), conceptual change and misconceptions, procedures, habits, and skills, and correction (refutation texts, correction of misinformation, learning from errors), together with the definitions imported from organization studies. The period was 1942 to 2026. Japanese peer-reviewed articles were included. Organizational unlearning was used only to compare definitions, and discussions of design were not covered.
• Collection: 90 candidates were gathered across academic databases (OpenAlex, Crossref, J-STAGE), and Crossref was used to check for retractions, corrections, and predatory journals (0 retractions, 0 suspected predatory journals, 2 corrections or addenda).
• Verification: of the 66 references, the full texts of 30 were obtained (including author versions, institutional repositories, PMC, and J-STAGE), and numbers and claims were checked against the relevant passages. For 29, only the abstract was available, and their content is reported as the authors’ claims. For 7, neither the full text nor the abstract could be obtained; their content is reported through the descriptions in other sources whose full texts were read, and this is stated in the text.
• Reporting: for each study, the note states whether the old content “disappeared” or “became less likely to be used,” together with how this was measured (recall performance, response times, speed of relearning, or behavior).

How has “unlearning” been defined?

The term unlearning spread to its current extent from organization studies. According to their abstract, Tsang and Zahra (2008) noted that organizational unlearning is widely considered a condition for adapting to environmental change but has received little research attention, and they discussed its definition, its distinction from organizational learning, and its dimensions. Klammer and Gueldenberg (2019) systematically reviewed 63 works on organizational unlearning and forgetting and wrote that robust conceptual and empirical evidence is still needed (abstract). They proposed a framework that positions studies on two axes: whether the loss of knowledge is intentional and how deep the loss is.

At the individual level, definitions diverge further. Hislop, Bosley, Coombs, and Holland (2014) noted that individual unlearning has received little attention, and, based on a conceptual review, distinguished two types of individual unlearning and developed a typology of four types (abstract). In health care, Rushmer and Davies (2004) distinguished “routine unlearning,” which merely requires establishing new habits for everyday procedures, from “deep unlearning,” which requires a substantive break with previous modes of understanding and doing, and described the latter as sudden and potentially painful (abstract). In nursing education, Macdonald (2002) used her own experience of giving up a trusted nursing practice concerning infant sleep positions to describe unlearning as a process of personal growth that involves receptivity, recognition, and grieving (abstract). These definitions describe letting go as discarding or abandoning knowledge or practices.

This assumption has been challenged from the side of psychology. Howells and Scholderer (2016) argued that the definition of unlearning that organization studies claims to have imported from psychology through Hedberg, namely the manageable discard of knowledge preceding and aiding later learning, is not supported by the psychology literature (abstract). When they re-examined highly cited empirical studies in organization studies, what had been presented under the label of unlearning was a conventional process of theory change, in which an established understanding is set aside (not deleted) in favor of a new one when new facts are presented. They recommended existing concepts such as learning, theory change, discard of practice, and extinction instead. Responding to this skepticism that knowledge cannot be discarded, Grisold, Kaiser, and Hafner (2017) redefined individual unlearning as reducing the influence of old knowledge on cognition (abstract). This raises the question of what happens to old knowledge itself when its influence is reduced.

Do old memories disappear?

In memory research, the idea that old learning is erased by new learning was once part of interference theory. According to Anderson, Bjork, and Bjork (1994), the classical two-factor theory of interference included “unlearning,” the weakening of old associations while new pairs are learned, as one of the mechanisms of forgetting (p.1081). In a representative experiment by Barnes and Underwood (1959), recall of responses from the first list decreased as learning trials on the second list increased (also as described by Anderson et al. 1994). Anderson et al. reinterpreted this result and suggested that practice in retrieving the second list from cues may have suppressed the competing first-list responses. A decrease in recall alone cannot distinguish whether the old memory disappeared or lost out at the stage of retrieval.

The new theory of disuse by Bjork and Bjork expresses this distinction with two indices of memory strength. Storage strength represents how well a memory is stored, and retrieval strength represents how easily it can be retrieved at present; when a memory is not used for a long period, retrieval strength declines even while storage strength remains high (as summarized by Kluge and Gronau 2018). In this theory, long-term memory has no capacity limit, and forgetting is mainly a decline in retrieval strength. Bjork and Bjork (2019) stated that conditions that produce forgetting, such as changing the context, increasing the delay, and interleaving practice, also enhance learning when the material is restudied (abstract).

Retrieval makes other memories harder to recall. In three experiments, Anderson et al. (1994) showed that practicing the retrieval of some words in a category impaired later recall of unpracticed words from the same category on a delayed test, and called this retrieval-induced forgetting. Murayama, Miyatsu, Buchli, and Storm (2014) meta-analyzed 512 samples from 143 papers and reported an effect size of g = 0.35 (95% confidence interval 0.32 to 0.38), which corresponds to a difference in recall of about 9%. Studies that controlled interference arising from the order of output at test showed a smaller effect (g = 0.22, compared with 0.50 in studies without this control), and studies with text materials that applied the same control showed g = 0.08, which was not significant. With text materials, practice on some parts sometimes facilitated recall of the unpracticed parts, and the authors considered that the integration of the content determined the effect. On tests given 24 hours or more later, the effect remained when text materials were excluded (g = 0.40 across 21 samples). The authors concluded that the results largely supported inhibition-based accounts but also included challenging evidence.

Trying not to retrieve a memory also produces forgetting. The Think/No-Think procedure, introduced by Anderson and Green (2001), asks participants to learn cue-target pairs and then, for some cues, to prevent the associated target from coming to mind (as described by Stramaccia et al. 2021). In their abstract, Clark et al. (2026) reported a meta-analysis of 500 effects from 120 studies, in which this suppression-induced forgetting was consistently observed, with effect sizes of d = 0.20 to 0.40, or about 6% when measured with the same cues. The effect varied considerably across studies, and when measured with the same cues it was larger with more suppression repetitions and with instructions to substitute another thought. In a meta-analysis of 25 studies, Stramaccia et al. (2021) wrote that healthy participants showed small-to-moderate forgetting, that the effect was weaker in anxious and depressed individuals, and that the estimate for healthy participants may have been inflated by publication bias (abstract).

Both retrieval-induced forgetting and suppression-induced forgetting measure how difficult retrieval has become. A different measure is needed to examine whether a memory has disappeared. Research on conditioning has such a measure. Bouton (2004) reviewed evidence that extinction does not destroy the original learning but instead generates new learning that is especially dependent on context (abstract). Spontaneous recovery, in which an extinguished response returns over time, renewal, in which it returns when the context is changed, and faster reacquisition when rewards are reintroduced have been treated as evidence that the original learning remains (as summarized by Crossley et al. 2013).

Can a memory then be rewritten when it is reactivated? Nader, Schafe, and LeDoux (2000) showed that a consolidated fear memory in rats returns to a temporarily unstable state when it is reactivated, and that blocking protein synthesis at that time leads to the loss of the memory (as described by Elsey et al. 2018). Studies of this reconsolidation in human memory followed. In Hupbach, Gomez, Hardt, and Nadel (2007), college students who learned a list of objects on day 1, received a reminder on day 2, and then learned a second list, intermixed objects from the second list when recalling the first list on day 3. The number of items recalled from the first list did not decrease; the memory did not disappear but changed by incorporating new information. Schiller et al. (2010) reported that, in a group that received a single presentation of the cue to reactivate a fear memory on the day after conditioning and underwent extinction training 10 minutes later, spontaneous recovery of the fear response did not occur, and that the effect lasted at least a year. Recovery occurred in the group that underwent extinction 6 hours after the reminder and in the group that was not reminded.

However, these results have not replicated consistently. Hardwicke, Taqi, and Shanks (2016) made four direct replication attempts (64 participants) of a motor-sequence learning study that is widely cited as evidence of human reconsolidation and did not observe the memory impairment shown in the original study. In three conceptual replications (48 participants) with a recall task using sequences similar to telephone numbers, retrieval helped preserve existing memories rather than updating them. Elsey, Van Ast, and Kindt (2018) reviewed studies of human reconsolidation and concluded that, although many observations are consistent with reconsolidation, inconsistent findings and alternative explanations mean that reconsolidation at the neurobiological level cannot be conclusively inferred from current evidence. They also cited a study in which another laboratory could not replicate the basic finding of Hupbach et al. and studies in which reactivation followed by extinction was not superior to standard extinction.

What memory research shows is that old memories becoming harder to retrieve is better established than old memories disappearing. Difficulty of retrieval can be produced by practicing the retrieval of other items, by trying not to retrieve, and by periods of disuse. In studies that tested whether memories had disappeared, old responses repeatedly returned with context and time. The reconsolidation procedures, which produced the results closest to erasure, have not replicated reliably in humans.

Does learning science replace naive ideas?

In school science and mathematics, the old content to be let go corresponds to the naive ideas that learners hold before instruction (misconceptions, naive conceptions). The most cited model in conceptual change research, that of Posner et al. (1982), proposed that an old conception is replaced by a new one (accommodation) when learners become dissatisfied with the old conception and find the new one intelligible, plausible, and fruitful. From this model, cognitive conflict, which makes learners aware of inconsistencies between the old and new conceptions, was derived as the central instructional strategy (as summarized by Paçacı et al. 2024, pp.4–5).

Later theories diverged according to what naive ideas are taken to be (Murayama 2011). Vosniadou’s framework theory regards naive ideas as coherent, theory-like structures supported by presuppositions formed in early childhood, and addresses how taught scientific ideas combine with naive presuppositions. Chi (2005) argued that misconceptions of emergent processes such as diffusion are especially robust because learners classify them as the same kind of process as direct processes such as blood circulation, and that correcting them requires re-representation across ontological kinds (abstract). diSessa’s knowledge in pieces regards naive ideas as collections of small knowledge elements (p-prims) abstracted from everyday experience. In Murayama’s summary of this view, a p-prim that is no longer used has not been replaced by another p-prim; it has only been removed from the cueing relation to the phenomenon, and the element itself remains and can be used on another occasion (Murayama 2011, pp.225–226). Smith, diSessa, and Roschelle (1994) criticized the view that misconceptions must be confronted and replaced, regarded learners’ prior conceptions as resources for developing expertise, and argued that refinement and reorganization, rather than replacement, should be the primary metaphors for learning (abstract).

Some studies have examined whether replacement occurred by measuring response times. Shtulman and Valcarcel (2012) asked 150 college students to verify, as quickly as possible, 200 statements in 10 domains: astronomy, evolution, fractions, genetics, germs, matter, mechanics, physiology, thermodynamics, and waves. The statements consisted of those whose truth value is the same on naive and scientific theories (“The moon revolves around the Earth”) and those whose truth value reverses between the two theories (“The Earth revolves around the sun”). Statements whose truth value reverses were verified more slowly and less accurately in all 10 domains. In fractions, germs, and physiology, where overall accuracy was 80% or higher, statements whose truth value was the same were verified 562 ms faster on average. In 6 domains, participants with higher accuracy showed a larger delay on statements whose truth value reverses. The authors concluded that naive theories are suppressed by scientific theories but not supplanted by them.

This delay remains long after learning and among experts. Shtulman and Harrington (2016) gave the same task to 104 younger adults and 48 older adults (mean age 65.1 years) and reported that older adults were as accurate as younger adults but showed a larger delay on statements whose truth value reverses, and that the delay persisted among professional scientists. Twenty-five scientists with a doctorate in physics were also slower and less accurate when judging the scientific value of statements corresponding to naive ideas than control statements. At that time, a network of frontal brain regions associated with inhibitory control was more active (Allaire-Duquette et al. 2021). Masson et al. (2014) showed that, when judging whether a circuit in which a bulb is connected to a battery by a single wire is correct, experts activated brain regions associated with inhibition more than novices did (abstract). In Kelemen, Rottman, and Seston (2013), physical scientists from top-ranked American universities more often judged unwarranted teleological explanations of natural phenomena as true when they were under time pressure (abstract). The scientists endorsed such explanations less than the comparison groups did, but they did not differ from humanities scholars.

However, the evidence from response times has been qualified from the side of measurement. Potvin, Masson, Lafortune, and Cyr (2015) pointed out that interpreting response-time delays as evidence of coexistence depends on the assumption that stimuli in different conditions differ only in the presence or absence of an interfering misconception (abstract). They designed a task that varied the intensity of interference in steps for the misconception that “heavier objects sink more,” and showed in 128 students aged 14 to 15 that stronger interference produced longer delays, which supported the coexistence hypothesis. Edelsbrunner (2025) examined individual differences in the task of Shtulman and Valcarcel using factor analysis and showed that there is little stable variation in individuals’ increases in response times and error rates, either across domains or across topics within domains (abstract). The author suggested that the size of interference may depend on content knowledge and context rather than on stable traits such as inhibition ability. The persistence of old ideas has been observed repeatedly, but the way they persist cannot yet be measured as a stable trait of individuals.

Classroom intervention studies have also begun to assume a shift in prevalence rather than replacement. Potvin, Sauriol, and Riopel (2015) compared three sequences with 558 students in grades 5 and 6: a classical sequence that triggers conflict first, a sequence that presents the scientific explanation first and then triggers conflict, and a sequence that repeats the explanation without triggering conflict (abstract). The second sequence was based on the prevalence model, in which several conceptions coexist and one of them comes to prevail over the others. The authors wrote that conflict was useful for conceptual change but did not need to be triggered at the beginning, that the prevalence model might be preferable for use in schools, and that some results showed small effect sizes.

In Japan, Shindo, Makara, and Fushimi (2006) proposed the “fusion method.” This method first presents a problem that cannot be solved by applying the misconception and then a similar problem that can be solved even with the misconception, and it uses the correct solution of the latter to help learners grasp the correct rule and apply it to a series of similar problems. In an experiment with 76 university students, the fusion method corrected the misconception (that a vacuum pulls objects toward it) better than a method using counterexamples and a method not using them, and also increased learners’ awareness that their knowledge had changed and their interest. Nakanishi, Ohmichi, and Umemoto (2018) reported, with 135 university and junior college students, that a confrontational strategy, which begins with information that conflicts with existing concepts, promoted conceptual change more than a conciliatory strategy, which begins with information consistent with them. Sato and Nagayama (2025) showed that the effect of a confrontational strategy differed according to how learners had answered before instruction. Of 46 university students, those who consistently gave the same wrong answer before instruction became able to apply the rule after instruction, whereas those who judged each problem individually still could not solve the problems afterward, and interviews indicated that they had difficulty constructing knowledge that linked cases with the rule. This result suggests that effective instruction may differ depending on whether naive ideas form a coherent, theory-like structure or a collection of fragments.

Are established procedures and habits rewritten?

Old content also interferes with new content at the level of procedures. In the water-jar problems of Luchins (1942), participants first solved five problems with the same long procedure and then received problems that could also be solved with a shorter procedure. More than 80% of participants did not notice the shorter procedure and continued to use the longer one (as described by Bilalić et al. 2008a). Bilalić, McLeod, and Gobet (2008a) used chess problems that had both a familiar but non-optimal solution and a better but less familiar one, and showed that this Einstellung effect also occurred in experts, reducing their problem-solving ability to about that of players three standard deviations lower in skill. However, the more expert the players were, the less prone they were to the effect. In a study by the same authors that measured eye movements (2008b), experts who had found the first solution reported that they were looking for a better one, but their eyes kept returning to the features of the board related to the first solution. The old procedure determined where attention went during the search for a new one.

Expertise itself can also reverse the effectiveness of instruction. Kalyuga, Ayres, Chandler, and Sweller (2003) reviewed empirical studies to show the expertise reversal effect, in which instructional techniques that are effective for novices lose their effectiveness, and sometimes have negative consequences, for more experienced learners (abstract). Dane (2010) reconsidered the view that flexibility is lost with expertise as cognitive entrenchment, a high level of stability in one’s domain schemas, and argued that the degree of entrenchment varies not only with expertise but also with the task environment and attentional focus (abstract).

Habits are difficult to change even when goals change. Wood and Neal (2007) described habits as direct associations between contexts and responses that accrue slowly, so that a context can trigger a response without the mediation of a goal. In an experiment by Betsch and colleagues cited by Wood and Neal, participants first formed habits for subway routes, and the correct routes were then switched. Participants formed implementation intentions that linked the new routes to cues and were paid according to performance; nevertheless, they chose the previous routes in about 50% of their responses (p.855). Wood and Neal identified two ways of changing habits. The first is to change exposure to habit cues, either through deliberate avoidance or through changes in life circumstances such as moving house or changing jobs (p.859). In a study of students transferring to a new university, among students with strong habits at the old university, those who reported that the locations or partners of activities such as exercise had changed did not maintain their habits and acted in line with their goals at the new university. Those whose contexts had not changed maintained their habits regardless of their goals (p.856). The second is to inhibit the response after the cue has appeared; an avoidance strategy of vigilantly monitoring for the unwanted response was especially effective, but inhibition is effortful and draws on a limited self-regulatory resource (p.859).

Planning methods are also less effective when habits are strong. Webb, Sheeran, and Luszczynska (2009) showed that if-then plans (implementation intentions) reduced the extent to which habits impaired performance after a task switch, but that the effect was smaller when habits were stronger (abstract). For smoking, the plans reduced smoking among participants with weak or moderate habits but not among those with strong habits. New habits also take time to develop: in Lally et al. (2010), the time it took for the automaticity of a behavior performed daily in the same context to reach 95% of its asymptote ranged from 18 to 254 days across individuals (abstract).

For motor and procedural skills, mechanisms by which old learning is protected rather than erased have been examined. Krakauer, Ghez, and Ghilardi (2005) examined interference in which learning to adapt to a 30° rotation of a screen cursor no longer produced faster relearning the next day when a counter-rotation was learned 5 minutes after the original rotation. When washout blocks were introduced before learning, resistance to interference was observed when the counter-rotation was learned 24 hours later. Crossley, Ashby, and Maddox (2013) explained the difficulty of erasing procedural skills such as a golf swing with a computational model of striatal learning. When rewards are no longer contingent on behavior, striatal interneurons that act as a gate for learning stop pausing, and learning is thereby protected from decay. The model predicted that making rewards partially contingent on behavior would leave room for unlearning, and three behavioral experiments supported this prediction. They required, as a condition for calling a change unlearning, that relearning proceed no faster than the original learning when the original training conditions are reintroduced, in addition to the disappearance of the behavior. This is because relearning is faster when the old skill remains. In sport, Carson and Collins (2011) pointed out that research on technical change in skilled athletes is scarce and proposed, based on the coaching and psychology literature, a five-stage model for making such changes secure under competitive pressure (abstract).

What helps people switch from old knowledge?

If old knowledge cannot be erased, what instruction can aim for is that new knowledge wins in situations where old knowledge is activated. A refutation text is a tool that puts this aim into the form of a text: it states a common misconception, refutes it, and explains the correct conception (Schroeder and Kucera 2022). Using disruption during reading as an indicator, Kendeou, Walsh, Smith, and O’Brien (2014) reported that texts combining refutation and explanation reduced the disruption caused by commonsense beliefs, that refutation alone reduced but did not eliminate the disruption, and that explanation alone was as effective as the combination (abstract). In the fourth experiment, the combination of refutation and explanation also produced long-term learning outcomes. They explained these results with the Knowledge Revision Components (KReC) framework. In this explanation, stating the misconception activates existing knowledge, and immediately presenting the correct explanation draws activation away from the misconception and toward the correct idea (as described by Schroeder and Kucera 2022).

The effect of refutation texts has been confirmed repeatedly in meta-analyses. Guzzetti et al. (1993) concluded that texts can be used to eliminate misconceptions when they are refutational or when they are combined with other strategies that produce cognitive conflict (as quoted by Schroeder and Kucera 2022). In the review of 22 studies by Tippett (2010), the 10 studies of grades 3 to 10 found that refutation texts were more likely than expository texts to produce conceptual change, whereas one study of kindergarten to grade 2 and 11 studies of grade 11 and above found no significant benefit (also as described by Schroeder and Kucera 2022). In a meta-analysis of 44 independent comparisons (3,869 participants), Schroeder and Kucera (2022) found a moderate effect (g = 0.41), which did not vary with learners’ age, domain, type of test, or timing of the test. However, only two studies tested learners more than one month later. In a pre-registered meta-analysis of 71 articles (76 studies, 294 effect sizes), Danielson et al. (2025) showed that refutation texts increased accurate beliefs (g = 0.37) and reduced inaccurate beliefs (g = −0.38). None of the 26 moderators enhanced or diminished the effect. Effects on emotions and attitudes did not differ from zero.

Nevertheless, evidence on how long the effects last and whether they change behavior is limited. Zengilowski, Schuetze, Nash, and Schallert (2021) identified methodological problems in refutation text research: the use of a single topic or very few topics per study, the role of testing itself, and the durability of change beyond an immediate posttest (abstract). Ferrero, Hardwicke, Konstantinidis, and Vadillo (2020) had teachers in Spain read texts refuting misconceptions about education (such as those concerning learning styles). Belief in the refuted ideas decreased, but in the second experiment, in which beliefs were measured again 45 days later, the effect was no longer detected, and intentions to implement practices based on the misconceptions did not decrease. Ratings of intention were, if anything, higher after reading the refutations. For conceptual change strategies in general, Paçacı, Üstün, and Özdemir (2024) reported a large effect (g = 1.10) across 218 studies (18,051 students), but the prediction interval was wide (0.19 to 2.38), the outcome was mainly science achievement, and comparisons of delayed effects were left for future studies. In a review of biology interventions, Aleknavičiūtė, Lehtinen, and Södervik (2023) wrote that effect sizes were strongly influenced by sample size and publication bias, and that many studies reported only superficial learning outcomes, such as knowledge enrichment rather than knowledge restructuring (abstract).

Research on correcting misinformation also shows that old information remains after correction. Lewandowsky, Ecker, Seifert, Schwarz, and Cook (2012) summarized experiments on retractions and wrote that retractions rarely, if ever, eliminate reliance on misinformation, even when people believe, understand, and later remember the retraction (the continued influence effect). In a meta-analysis of 52 studies (6,878 participants), Chan, Jones, Hall Jamieson, and Albarracín (2017) showed that debunking had a large effect (d = 1.14 to 1.33), whereas the persistence of misinformation after debunking was also large (d = 0.75 to 1.06). When recipients generated reasons supporting the initial misinformation, persistence was stronger and the debunking effect was weaker. Walter and Murphy (2018) reported across 65 studies that corrective messages moderately reduced belief in misinformation (r = .35), that real-world misinformation was harder to correct than misinformation constructed for experiments (r = .14 and .48), and that appeals to coherence were more effective than fact-checking (r = .55 and .25) (abstract). Focusing on science-relevant misinformation, Chan and Albarracín (2023) found across 245 effect sizes from 75 reports that the effect of corrections was, on average, not significant (d = 0.11, 95% confidence interval −0.04 to 0.26; values after the 2025 correction). Corrections were more effective when the issue was not politically polarized and when recipients were already familiar with both sides of the issue.

When people correct their own errors, the strength of confidence can work in the opposite direction. Butterfield and Metcalfe (2001) found that errors committed with high confidence were more likely than low-confidence errors to be corrected on a retest after the correct response was shown (abstract). Metcalfe (2017) summarized that errors followed by corrective feedback are beneficial for learning and that correction should include an analysis of the reasoning that led to the error (abstract). These results were measured by whether participants chose the correct answer on a retest; they did not measure whether the old answer remained as a delay in response times. The lineage of designs that build errors into learning is examined in Designing Failure Into Learning: From Expectation Failure to Productive Failure and Industry Implementations.

How can letting go be verified?

The studies above measured whether old content had been let go with different measures.

Individual unlearning in workplaces and health care has been examined mainly through interviews. Based on interviews with 15 primary care physicians at a U.S. Veterans Affairs medical center, Gupta, Boland, and Aron (2017) reported that abandoning an outmoded clinical practice is a struggle amid tensions between evidence and context, that change is not unidirectional, and that clinical practice is in constant flux. Hafner (2019) conducted three weekly interviews with 31 participants who had experienced updates of procedures, classified their unlearning using the typology of Rushmer and Davies, and found two predominant types of knowledge change (abstract). Kato and Ozawa (2022) analyzed semi-structured interviews with seven third-year university students using the trajectory equifinality approach and described a process in which experiences of moving between lecture-centered classes and seminar or project classes led to unlearning of the value of “being forced to cram knowledge.” Kluge (2023) reviewed 31 empirical studies on organizational unlearning and intentional forgetting from 2019 to 2022, stated that many used static cross-sectional designs that do not allow causal conclusions, and wrote that longitudinal designs that follow the process are needed. The condition that Crossley et al. (2013) set for procedural skills (the behavior disappears, and relearning is no faster than the original learning) is one of the few operational definitions that use behavior to verify that old learning no longer remains.

Which claims does the literature support, and which does it not?

The way of letting go that the literature describes is not erasing the old but arranging conditions under which the new prevails while the old remains. The conditions that the literature identified were activating the old idea and then presenting the correct explanation (Kendeou et al. 2014) and changing exposure to cues (Wood and Neal 2007). The ways of verification that were recommended were measuring again after a delay rather than relying on an immediate test, and following the same people or organizations over time (Zengilowski et al. 2021; Kluge 2023).

Gaps

• Within the range searched, few studies followed the same people from before an intervention over several months while measuring both how old knowledge remains (response times, speed of relearning) and how new knowledge is used. Studies of organizations and workplaces mostly use cross-sectional designs (Kluge 2023).
• No intervention study was found that measured the switch from old skills or knowledge itself in vocational training or reskilling.
• Few refutation text studies measured outcomes more than one month later (two in the studies analyzed by Schroeder and Kucera 2022), and even fewer measured behavior.
• Rewriting human memory through reconsolidation has not replicated reliably and is not at a stage where its use in education can be examined.
• The intrusion of old ideas measured by response times has not been measured as a stable difference between individuals (Edelsbrunner 2025).

Unverified items

• No major claim remains unverified.
• For 29 sources, the full text could not be obtained and only the abstract was used (Tsang and Zahra 2008, Klammer and Gueldenberg 2019, Hislop et al. 2014, Rushmer and Davies 2004, Macdonald 2002, Howells and Scholderer 2016, Grisold et al. 2017, Bjork and Bjork 2019, Clark et al. 2026, Bouton 2004, Chi 2005, Smith et al. 1994, Masson et al. 2014, Kelemen et al. 2013, the two papers by Potvin et al. 2015, Edelsbrunner 2025, Kalyuga et al. 2003, Dane 2010, Webb et al. 2009, Lally et al. 2010, Carson and Collins 2011, Kendeou et al. 2014, Zengilowski et al. 2021, Aleknavičiūtė et al. 2023, Walter and Murphy 2018, Butterfield and Metcalfe 2001, Metcalfe 2017, Hafner 2019). In the text, their content is attributed to the authors with “abstract.”
• For 7 sources (Barnes and Underwood 1959, Anderson and Green 2001, Nader et al. 2000, Posner et al. 1982, Luchins 1942, Guzzetti et al. 1993, Tippett 2010), neither the full text nor the abstract could be obtained; their content is reported through the descriptions in sources whose full texts were read (Anderson et al. 1994, Stramaccia et al. 2021, Elsey et al. 2018, Paçacı et al. 2024, Bilalić et al. 2008a, Schroeder and Kucera 2022), and the source is stated in the text.
• Stramaccia et al. (2021) was read in a preprint version, and no numbers from it are reported. Murayama et al. (2014) was read in the accepted manuscript, and Elsey et al. (2018) in the version included in the first author’s doctoral thesis.
• Chan and Albarracín (2023) was read in the author manuscript that reflects the 2025 correction, and the numbers were checked against the correction notice.

References

All sources were accessed on 2026-11-13.

Definitions and imports from organization studies

• Tsang, E. W. K., & Zahra, S. A. (2008). Organizational unlearning. Human Relations, 61(10), 1435–1462. https://doi.org/10.1177/0018726708095710
• Klammer, A., & Gueldenberg, S. (2019). Unlearning and forgetting in organizations: A systematic review of literature. Journal of Knowledge Management, 23(5), 860–888. https://doi.org/10.1108/JKM-05-2018-0277
• Hislop, D., Bosley, S., Coombs, C. R., & Holland, J. (2014). The process of individual unlearning: A neglected topic in an under-researched field. Management Learning, 45(5), 540–560. https://doi.org/10.1177/1350507613486423
• Rushmer, R., & Davies, H. T. O. (2004). Unlearning in health care. Quality and Safety in Health Care, 13(Suppl 2), ii10–ii15. https://doi.org/10.1136/qshc.2003.009506
• Macdonald, G. (2002). Transformative unlearning: Safety, discernment and communities of learning. Nursing Inquiry, 9(3), 170–178. https://doi.org/10.1046/j.1440-1800.2002.00150.x
• Howells, J., & Scholderer, J. (2016). Forget unlearning? How an empirically unwarranted concept from psychology was imported to flourish in management and organisation studies. Management Learning, 47(4), 443–463. https://doi.org/10.1177/1350507615624079
• Grisold, T., Kaiser, A., & Hafner, J. (2017). Unlearning before creating new knowledge: A cognitive process. In Proceedings of the 50th Hawaii International Conference on System Sciences (pp. 4614–4623). https://doi.org/10.24251/HICSS.2017.561

Memory

• Barnes, J. M., & Underwood, B. J. (1959). “Fate” of first-list associations in transfer theory. Journal of Experimental Psychology, 58(2), 97–105. https://doi.org/10.1037/h0047507
• Anderson, M. C., Bjork, R. A., & Bjork, E. L. (1994). Remembering can cause forgetting: Retrieval dynamics in long-term memory. Journal of Experimental Psychology: Learning, Memory, and Cognition, 20(5), 1063–1087. https://doi.org/10.1037/0278-7393.20.5.1063
• Murayama, K., Miyatsu, T., Buchli, D., & Storm, B. C. (2014). Forgetting as a consequence of retrieval: A meta-analytic review of retrieval-induced forgetting. Psychological Bulletin, 140(5), 1383–1409. https://doi.org/10.1037/a0037505
• Bjork, R. A., & Bjork, E. L. (2019). Forgetting as the friend of learning: Implications for teaching and self-regulated learning. Advances in Physiology Education, 43(2), 164–167. https://doi.org/10.1152/advan.00001.2019
• Kluge, A., & Gronau, N. (2018). Intentional forgetting in organizations: The importance of eliminating retrieval cues for implementing new routines. Frontiers in Psychology, 9, 51. https://doi.org/10.3389/fpsyg.2018.00051
• Anderson, M. C., & Green, C. (2001). Suppressing unwanted memories by executive control. Nature, 410(6826), 366–369. https://doi.org/10.1038/35066572
• Clark, C. A., Todorovic, D., Levy, B. J., Eschmann, K. C. J., Klein, M. D., Anderson, M. C., van Schie, K., Whitridge, J. W., Yick, J. J. J., & Fawcett, J. M. (2026). Forgetting as a consequence of retrieval suppression: A meta-analytic review. Journal of Experimental Psychology: General, 155(5), 1318–1346. https://doi.org/10.1037/xge0001922
• Stramaccia, D. F., Meyer, A.-K., Rischer, K. M., Fawcett, J. M., & Benoit, R. G. (2021). Memory suppression and its deficiency in psychological disorders: A focused meta-analysis. Journal of Experimental Psychology: General, 150(5), 828–850. https://doi.org/10.1037/xge0000971
• Bouton, M. E. (2004). Context and behavioral processes in extinction. Learning & Memory, 11(5), 485–494. https://doi.org/10.1101/lm.78804
• Nader, K., Schafe, G. E., & LeDoux, J. E. (2000). Fear memories require protein synthesis in the amygdala for reconsolidation after retrieval. Nature, 406(6797), 722–726. https://doi.org/10.1038/35021052
• Hupbach, A., Gomez, R., Hardt, O., & Nadel, L. (2007). Reconsolidation of episodic memories: A subtle reminder triggers integration of new information. Learning & Memory, 14(1–2), 47–53. https://doi.org/10.1101/lm.365707
• Schiller, D., Monfils, M.-H., Raio, C. M., Johnson, D. C., LeDoux, J. E., & Phelps, E. A. (2010). Preventing the return of fear in humans using reconsolidation update mechanisms. Nature, 463(7277), 49–53. https://doi.org/10.1038/nature08637
• Hardwicke, T. E., Taqi, M., & Shanks, D. R. (2016). Postretrieval new learning does not reliably induce human memory updating via reconsolidation. Proceedings of the National Academy of Sciences, 113(19), 5206–5211. https://doi.org/10.1073/pnas.1601440113
• Elsey, J. W. B., Van Ast, V. A., & Kindt, M. (2018). Human memory reconsolidation: A guiding framework and critical review of the evidence. Psychological Bulletin, 144(8), 797–848. https://doi.org/10.1037/bul0000152

Conceptual change and naive ideas

• Posner, G. J., Strike, K. A., Hewson, P. W., & Gertzog, W. A. (1982). Accommodation of a scientific conception: Toward a theory of conceptual change. Science Education, 66(2), 211–227. https://doi.org/10.1002/sce.3730660207
• Chi, M. T. H. (2005). Commonsense conceptions of emergent processes: Why some misconceptions are robust. Journal of the Learning Sciences, 14(2), 161–199. https://doi.org/10.1207/s15327809jls1402_1
• Smith, J. P., III, diSessa, A. A., & Roschelle, J. (1994). Misconceptions reconceived: A constructivist analysis of knowledge in transition. Journal of the Learning Sciences, 3(2), 115–163. https://doi.org/10.1207/s15327809jls0302_1
• Murayama, I. (2011). Theories of conceptual change (概念変化についての諸理論). Japanese Psychological Review (心理学評論), 54(3), 218–231. https://doi.org/10.24602/sjpr.54.3_218 (in Japanese)
• Shtulman, A., & Valcarcel, J. (2012). Scientific knowledge suppresses but does not supplant earlier intuitions. Cognition, 124(2), 209–215. https://doi.org/10.1016/j.cognition.2012.04.005
• Shtulman, A., & Harrington, K. (2016). Tensions between science and intuition across the lifespan. Topics in Cognitive Science, 8(1), 118–137. https://doi.org/10.1111/tops.12174
• Allaire-Duquette, G., Brault Foisy, L.-M., Potvin, P., Riopel, M., Larose, M., & Masson, S. (2021). An fMRI study of scientists with a Ph.D. in physics confronted with naive ideas in science. npj Science of Learning, 6, 11. https://doi.org/10.1038/s41539-021-00091-x
• Masson, S., Potvin, P., Riopel, M., & Brault Foisy, L.-M. (2014). Differences in brain activation between novices and experts in science during a task involving a common misconception in electricity. Mind, Brain, and Education, 8(1), 44–55. https://doi.org/10.1111/mbe.12043
• Kelemen, D., Rottman, J., & Seston, R. (2013). Professional physical scientists display tenacious teleological tendencies: Purpose-based reasoning as a cognitive default. Journal of Experimental Psychology: General, 142(4), 1074–1083. https://doi.org/10.1037/a0030399
• Potvin, P., Masson, S., Lafortune, S., & Cyr, G. (2015). Persistence of the intuitive conception that heavier objects sink more: A reaction time study with different levels of interference. International Journal of Science and Mathematics Education, 13(1), 21–43. https://doi.org/10.1007/s10763-014-9520-6
• Edelsbrunner, P. A. (2025). Does interference between intuitive conceptions and scientific concepts produce reliable inter-individual differences? A psychometric analysis. Science & Education, 34(4), 1971–1988. https://doi.org/10.1007/s11191-024-00500-8
• Potvin, P., Sauriol, É., & Riopel, M. (2015). Experimental evidence of the superiority of the prevalence model of conceptual change over the classical models and repetition. Journal of Research in Science Teaching, 52(8), 1082–1108. https://doi.org/10.1002/tea.21235
• Shindo, T., Makara, K., & Fushimi, Y. (2006). Effects of the “fusion method,” a strategy for using counterexamples to correct misconceptions (誤概念の修正に有効な反証事例の使用方略：「融合法」の効果). Japanese Journal of Educational Psychology (教育心理学研究), 54(2), 162–173. https://doi.org/10.5926/jjep1953.54.2_162 (in Japanese)
• Nakanishi, Y., Ohmichi, K., & Umemoto, T. (2018). Self-efficacy for the accuracy of knowledge and for knowledge reconstruction, and conceptual change (知識の正確性ならびに知識再構築に対する自己効力感と概念変化). Japanese Journal of Educational Psychology (教育心理学研究), 66(3), 199–211. https://doi.org/10.5926/jjep.66.199 (in Japanese)
• Sato, S., & Nagayama, T. (2025). Do “erroneous rules” interfere with learning scientific rules? An examination from the perspective of knowledge construction (「誤ルール」は科学的ルールの学習を妨害するのか：知識構成の観点からの検討). Japanese Journal of Educational Psychology (教育心理学研究), 73(2), 63–78. https://doi.org/10.5926/jjep.73.63 (in Japanese)

Procedures, habits, and skills

• Luchins, A. S. (1942). Mechanization in problem solving: The effect of Einstellung. Psychological Monographs, 54(6), i–95. https://doi.org/10.1037/h0093502
• Bilalić, M., McLeod, P., & Gobet, F. (2008a). Inflexibility of experts: Reality or myth? Quantifying the Einstellung effect in chess masters. Cognitive Psychology, 56(2), 73–102. https://doi.org/10.1016/j.cogpsych.2007.02.001
• Bilalić, M., McLeod, P., & Gobet, F. (2008b). Why good thoughts block better ones: The mechanism of the pernicious Einstellung (set) effect. Cognition, 108(3), 652–661. https://doi.org/10.1016/j.cognition.2008.05.005
• Kalyuga, S., Ayres, P., Chandler, P., & Sweller, J. (2003). The expertise reversal effect. Educational Psychologist, 38(1), 23–31. https://doi.org/10.1207/S15326985EP3801_4
• Dane, E. (2010). Reconsidering the trade-off between expertise and flexibility: A cognitive entrenchment perspective. Academy of Management Review, 35(4), 579–603. https://doi.org/10.5465/amr.35.4.zok579
• Wood, W., & Neal, D. T. (2007). A new look at habits and the habit–goal interface. Psychological Review, 114(4), 843–863. https://doi.org/10.1037/0033-295X.114.4.843
• Webb, T. L., Sheeran, P., & Luszczynska, A. (2009). Planning to break unwanted habits: Habit strength moderates implementation intention effects on behaviour change. British Journal of Social Psychology, 48(3), 507–523. https://doi.org/10.1348/014466608X370591
• Lally, P., van Jaarsveld, C. H. M., Potts, H. W. W., & Wardle, J. (2010). How are habits formed: Modelling habit formation in the real world. European Journal of Social Psychology, 40(6), 998–1009. https://doi.org/10.1002/ejsp.674
• Krakauer, J. W., Ghez, C., & Ghilardi, M. F. (2005). Adaptation to visuomotor transformations: Consolidation, interference, and forgetting. Journal of Neuroscience, 25(2), 473–478. https://doi.org/10.1523/JNEUROSCI.4218-04.2005
• Crossley, M. J., Ashby, F. G., & Maddox, W. T. (2013). Erasing the engram: The unlearning of procedural skills. Journal of Experimental Psychology: General, 142(3), 710–741. https://doi.org/10.1037/a0030059
• Carson, H. J., & Collins, D. (2011). Refining and regaining skills in fixation/diversification stage performers: The Five-A Model. International Review of Sport and Exercise Psychology, 4(2), 146–167. https://doi.org/10.1080/1750984X.2011.613682

Refutation, correction, and learning from errors

• Kendeou, P., Walsh, E. K., Smith, E. R., & O’Brien, E. J. (2014). Knowledge revision processes in refutation texts. Discourse Processes, 51(5–6), 374–397. https://doi.org/10.1080/0163853X.2014.913961
• Guzzetti, B. J., Snyder, T. E., Glass, G. V., & Gamas, W. S. (1993). Promoting conceptual change in science: A comparative meta-analysis of instructional interventions from reading education and science education. Reading Research Quarterly, 28(2), 116–159. https://doi.org/10.2307/747886
• Tippett, C. D. (2010). Refutation text in science education: A review of two decades of research. International Journal of Science and Mathematics Education, 8(6), 951–970. https://doi.org/10.1007/s10763-010-9203-x
• Schroeder, N. L., & Kucera, A. C. (2022). Refutation text facilitates learning: A meta-analysis of between-subjects experiments. Educational Psychology Review, 34(2), 957–987. https://doi.org/10.1007/s10648-021-09656-z
• Danielson, R. W., Jacobson, N. G., Patall, E. A., Sinatra, G. M., Adesope, O. O., Kennedy, A. A. U., Bhat, B. H., Ramazan, O., Akinrotimi, B., Nketah, G., Jin, G., & Sunday, O. J. (2025). The effectiveness of refutation text in confronting scientific misconceptions: A meta-analysis. Educational Psychologist, 60(1), 23–47. https://doi.org/10.1080/00461520.2024.2365628
• Zengilowski, A., Schuetze, B. A., Nash, B. L., & Schallert, D. L. (2021). A critical review of the refutation text literature: Methodological confounds, theoretical problems, and possible solutions. Educational Psychologist, 56(3), 175–195. https://doi.org/10.1080/00461520.2020.1861948
• Ferrero, M., Hardwicke, T. E., Konstantinidis, E., & Vadillo, M. A. (2020). The effectiveness of refutation texts to correct misconceptions among educators. Journal of Experimental Psychology: Applied, 26(3), 411–421. https://doi.org/10.1037/xap0000258
• Paçacı, C., Üstün, U., & Özdemir, Ö. F. (2024). Effectiveness of conceptual change strategies in science education: A meta-analysis. Journal of Research in Science Teaching, 61(6), 1263–1325. https://doi.org/10.1002/tea.21887
• Aleknavičiūtė, V., Lehtinen, E., & Södervik, I. (2023). Thirty years of conceptual change research in biology: A review and meta-analysis of intervention studies. Educational Research Review, 41, 100556. https://doi.org/10.1016/j.edurev.2023.100556
• Lewandowsky, S., Ecker, U. K. H., Seifert, C. M., Schwarz, N., & Cook, J. (2012). Misinformation and its correction: Continued influence and successful debiasing. Psychological Science in the Public Interest, 13(3), 106–131. https://doi.org/10.1177/1529100612451018
• Chan, M. S., Jones, C. R., Hall Jamieson, K., & Albarracín, D. (2017). Debunking: A meta-analysis of the psychological efficacy of messages countering misinformation. Psychological Science, 28(11), 1531–1546. https://doi.org/10.1177/0956797617714579
• Walter, N., & Murphy, S. T. (2018). How to unring the bell: A meta-analytic approach to correction of misinformation. Communication Monographs, 85(3), 423–441. https://doi.org/10.1080/03637751.2018.1467564
• Chan, M. S., & Albarracín, D. (2023). A meta-analysis of correction effects in science-relevant misinformation. Nature Human Behaviour, 7(9), 1514–1525. https://doi.org/10.1038/s41562-023-01623-8
• Butterfield, B., & Metcalfe, J. (2001). Errors committed with high confidence are hypercorrected. Journal of Experimental Psychology: Learning, Memory, and Cognition, 27(6), 1491–1494. https://doi.org/10.1037/0278-7393.27.6.1491
• Metcalfe, J. (2017). Learning from errors. Annual Review of Psychology, 68, 465–489. https://doi.org/10.1146/annurev-psych-010416-044022

Measurement and individual unlearning at work

• Gupta, D. M., Boland, R. J., Jr., & Aron, D. C. (2017). The physician’s experience of changing clinical practice: A struggle to unlearn. Implementation Science, 12, 28. https://doi.org/10.1186/s13012-017-0555-2
• Hafner, J. (2019). Are there typological characteristics of individual unlearning? In Proceedings of the 52nd Hawaii International Conference on System Sciences. https://doi.org/10.24251/HICSS.2019.680
• Kato, N., & Ozawa, S. (2022). The influence of learning experiences after entering university on university students’ conceptions of learning: Focusing on students at selective private universities in the Tokyo metropolitan area (大学入学後の学習経験が大学生の学習観に与える影響：首都圏の上位私立大学生を対象として). Japan Journal of Educational Technology (日本教育工学会論文誌), 46(3), 433–452. https://doi.org/10.15077/jjet.45135 (in Japanese)
• Kluge, A. (2023). Recent findings on organizational unlearning and intentional forgetting research (2019–2022). Frontiers in Psychology, 14, 1160173. https://doi.org/10.3389/fpsyg.2023.1160173

Related notes

• Is Expertise a One-Way Street? Unlearning and the Blind Spot of Reversible Mastery: an essay that questions the assumption that expertise develops in one direction only, through the hypothesis that unlearning is a process symmetrical to learning. It reads the Einstellung effect and the expertise reversal effect as material for the hypothesis. This note reports the same phenomena by limiting itself to what the literature measured.
• Learning by Practice with Knowledge Withheld, Learning by Knowledge with Practice Withheld: The Literature on Whether Withholding One Side Has Meaning: examines, in a discussion of whether either knowledge or practice should be withheld, how everyday experience forms counterintuitive misconceptions that are not corrected by adding observations (Chi 2005; Smith et al. 1994).
• Designing Failure Into Learning: From Expectation Failure to Productive Failure and Industry Implementations: examines the lineage of designs that build errors and impasses into learning (productive failure, desirable difficulties).
• Does the Quantity of Work and Practice Build Skill? Evidence on Repetitions, Case Volume, Years of Experience, Practice Scheduling, Working Hours, and Workplace Experience: examines reports in which performance declines with years of experience and environments in which experience teaches errors. It provides a view from the side of quantity on how old ways persist.
• What Does Design Ask People to Let Go Of, and What Does It Let Go Of? Unlearning Considered through a Broad View of Design: considers the same question from a broad view of design. It covers design fixation, design that changes practices, unmaking and non-use, and who decides what is let go of, and it uses the findings of this note (old knowledge remains, and changing the context works better for habits) as a bridge.