Sunday L&D Myths #12: You Can Help People UnLearn Things
I’m seeing unlearning pop up more and more. Conference talks, LinkedIn posts, thought leadership articles, vendor pitches; there’s a growing enthusiasm for the idea that before people can learn something new, we need to help them unlearn what they knew before. The premise sounds reasonable enough: people carry outdated knowledge, bad habits, and incorrect assumptions, which get in the way of adopting new practices. So we need to clear out the old before we can install the new.
I’ve written about this before, briefly, but the frequency with which I’m encountering it now, and the confidence with which it’s being presented as something L&D practitioners should actively design for, has prompted me to spend more time with the research literature. I wanted to make sure my thinking is still on solid ground. As it turns out, it is.
Before we dig into things, let’s clarify something that I know is going to come up: this isn’t an academic quibble about terminology. If we believe that unlearning is something we can facilitate, something we can design interventions around, we fundamentally misunderstand the challenge we face when helping people change how they do things in the workplace. We end up designing for a process that doesn’t exist, which means we’re not designing for the process that does. And that makes us less effective at the thing we’re here to do.
What does “unlearning” mean?
Before examining the evidence, it’s worth being clear about what people usually mean when they talk about unlearning in workplace contexts. The concept dates back to a 1981 book chapter by Bo Hedberg in the Handbook of Organisational Design, in which he defined unlearning as “a process through which learners discard knowledge” (Hedberg, 1981). The idea was that organisations, and by extension individuals, need to get rid of obsolete information before they can effectively acquire new knowledge. Think of it like emptying a cup before you can fill it with something else.
In L&D contexts, this has evolved into a practical claim: that we can design interventions to help people discard what they previously learned. Run an “unlearning session” to help people let go of old practices. Create activities to remove outdated knowledge. Treat the old information as something to be deleted so the new information has space to land.
The problem is that cognitive science, neuroscience, and memory research all tell us the same thing: this isn’t how brains work. At all.
Note: The above definition of unlearning may differ from those you’ve seen online. I’ve seen three or four, and I’ve tried to encapsulate them within what I’ve described above. This, in and of itself, should be a red flag. If everyone defines a term differently, then it’s highly likely they’re basing their definitions on very little actual information.
With this in mind, bear with me if this definition of unlearning doesn’t sit right with you, as from here on out we’re going to address the broader point rather than just this definition.
Brains update, they don’t delete
Decades of research into memory, learning, and knowledge revision reveal that the brain operates through additive and competitive mechanisms, not subtractive ones. When we appear to “unlearn” something, what’s happening is considerably more complex than deletion.
The most useful framework for understanding this comes from Robert and Elizabeth Bjork’s New Theory of Disuse, developed in 1992. The Bjorks distinguish between two indices of memory strength: storage strength and retrieval strength. Storage strength reflects how well-learned or entrenched a memory is; importantly, it only ever increases. Once you’ve learned something, that learning becomes more deeply embedded over time, not less. Retrieval strength, on the other hand, reflects how accessible that memory is at any given moment, and this fluctuates based on context, recency, and competition from other memories (Bjork and Bjork, 1992).
What this means is that something you cannot currently recall may still be “in there” with high storage strength but low retrieval strength. Your current performance depends entirely on retrieval strength, but the underlying memory hasn’t been erased. This explains why relearning is faster than initial learning, why old habits resurface in familiar contexts, and why ex-smokers still struggle at parties years after quitting. The old learning never disappeared; it just became harder to access.
Modern neuroscience has identified what researchers call “active forgetting” mechanisms. Michael Anderson and colleagues at Cambridge have shown that the brain has intrinsic neural processes that suppress memory accessibility, but these mechanisms reduce access to memories without destroying them. The preferred technical term is “active forgetting,” not un-learning, precisely because the evidence consistently shows that memories are not deleted but become inaccessible (Anderson et al., 2004).
Perhaps the most striking evidence comes from research on memory reconsolidation by Karim Nader at McGill University. When we retrieve a memory, it enters a temporarily labile, easy-to-alter state, allowing it to be modified. This sounds promising for un-learning advocates, until you examine what actually happens. The memory is updated and re-stabilised with modifications incorporated; the original memory is transformed, not deleted. As Lee, Nader, and Schiller wrote in Trends in Cognitive Sciences: “Many have suggested that such retrieval-induced plasticity is ideally placed to enable memories to be updated with new information” (Lee, Nader and Schiller, 2017). Note the word: updated. Not erased.
As a reader of the Instructional Design Tips Substack, you can get 25% off a ticket to the very first IDTX Evidence-Informed Practice Conference.
This one-day event is set for the 29th of May 2026 and will be held in Birmingham city centre, UK. The day will see us bring together researchers, scientists, and practitioners to discuss how we utilise the wealth of scientific understanding, research, and evidence to improve workplace training.
To claim your discounted ticket, head over to the IDTX website and use code CPDW25 at checkout.
Extinction learning: the clearest refutation
If you want to understand why unlearning doesn’t exist as a cognitive process, extinction learning provides the most compelling evidence. Mark Bouton at the University of Vermont has spent decades studying what happens when we try to “undo” learned associations, and his findings are unambiguous.
In classical conditioning terms, extinction occurs when you repeatedly present a conditioned stimulus without the unconditioned stimulus it was paired with. The conditioned response diminishes and eventually stops.
Consider a straightforward example.
Imagine you work somewhere with a particular notification sound that always precedes urgent, stressful requests from senior leadership. Over time, hearing that sound triggers an immediate stress response; your shoulders tense, your attention sharpens, you brace yourself.
That’s conditioning: the sound (conditioned stimulus or CS) has become associated with stressful demands (unconditioned stimulus or US), producing an automatic reaction (conditioned response or CR).
Now, suppose the system changes.
The same sound now precedes routine, low-stakes messages. After weeks of hearing it without the stressful follow-up, your reaction diminishes. Eventually, the sound barely registers. It looks like you’ve unlearned the stress response.
This looks like unlearning, but four phenomena prove that the original learning survives intact.
Spontaneous recovery: extinguished responses return over time. In rat studies, fear responses that appeared to be completely extinguished came back to full strength within days (Bouton, 2004).
Renewal: the extinguished response returns when tested in a different context from where extinction occurred.
Reinstatement: exposure to the unconditioned stimulus alone can restore the conditioned response.
Savings: relearning occurs faster than original learning, proving that the memory trace persists (Bouton, 2002).
Bouton concluded that: “Extinction does not destroy the first-learned information but instead reflects new learning” (Bouton, 2004).
During extinction, a new, competing memory is created. The CS-no US association is learned alongside the original CS-US association; it doesn’t replace it. Which memory gets expressed depends on context, recency, and other retrieval cues.
In other words, you now have two associations with that notification sound, not one. The original association (sound means stress) still exists. You’ve simply learned a second association alongside it (sound means routine). Both are stored.
Whether you feel anxious or calm when you hear the sound depends on which association your brain retrieves in that moment, and that depends on where you are, how recently each association has been activated, and what other cues are present. The old boardroom retrieves the stress association; your new desk retrieves the calm one. Neither has replaced the other.
The neural evidence supports this interpretation. Extinction involves the ventromedial prefrontal cortex providing top-down inhibition of amygdala-based responses. It’s suppression, not erasure. The original response remains intact in the amygdala, ready to emerge when conditions change.
What the psychology literature says
In a 2016 paper by John Howells and Joachim Scholderer in Management Learning, titled “Forget unlearning?”, a comprehensive review of both the psychology literature and the organisational learning literature that claims to study un-learning was conducted.
First, “unlearning” doesn’t appear in PsycINFO, the primary database indexing psychological research. If unlearning were a genuine cognitive process that psychologists studied, it would have a presence in the discipline’s main literature. It doesn’t. The term originated in organisational studies and was incorrectly attributed to psychology (Howells and Scholderer, 2016).
Second, when they examined highly-cited empirical articles in the management literature that claimed to study unlearning, they found that “none provide evidence of its existence.” What these studies actually showed was “a conventional process of theory-change, the setting aside (not deletion) of an established understanding in favour of new understanding when presented with perceived new facts” (Howells and Scholderer, 2016).
In other words, what people called unlearning was just learning. New information was acquired that competed with and sometimes dominated older information. Nothing was discarded.
Daniel Schacter’s framework of the Seven Sins of Memory provides another lens on this. Schacter categorises memory failures under headings like “transience” (weakening accessibility over time), “blocking” (retrieval thwarted despite intact storage), and “interference” (competing memories). Notably absent from his framework is anything resembling unlearning or deletion.
Schacter emphasises that memory is reconstructive rather than reproductive; memories are rebuilt each time from available fragments, influenced by current knowledge and context. The old information doesn’t get removed; it gets recontextualised (Schacter, 1999).
The conceptual change literature, which directly addresses how people revise existing knowledge and correct misconceptions, tells the same story. Michelene Chi’s framework identifies three types of conceptual change: belief revision, mental model transformation, and categorical shift. All involve reorganising knowledge structures, not removing them. Stella Vosniadou’s research on how children’s understanding of the Earth changes from flat to spherical shows progressive revision; the earlier misconceptions aren’t erased but gradually revised through framework theory change (Chi, 2008; Vosniadou, 2013).
Why this matters for our practice
Here’s where this stops being a debate about terminology and starts being a practical concern about how we design interventions.
If you believe unlearning is possible, you might design training that asks people to “let go” of old knowledge before introducing new practices. You might create activities focused on discarding previous assumptions. You might frame resistance to change as a failure to unlearn. And you’d be designing for a cognitive process that doesn’t exist.
What actually happens when someone appears to change their practice is this:
They acquire new knowledge that competes with the old.
If the new knowledge is practised sufficiently, in relevant contexts, with adequate feedback, it achieves higher retrieval strength than the older information.
The new knowledge comes to dominate behaviour in the situations where it matters.
But the old knowledge doesn’t go away. It remains stored, often with increasing storage strength through each failed retrieval attempt, waiting to resurface when conditions favour it.
This has several practical implications.
Expect interference, not erasure.
When people revert to old practices, they haven’t failed to unlearn. They’ve experienced normal retrieval competition where the old, more practised response won. The solution isn’t more attempts at erasure; it’s more practice of the new behaviour in conditions that match where it needs to be applied.
Context matters enormously.
Because different contexts activate different knowledge, ensure new learning occurs in contexts resembling where it will be applied. Transfer depends on context matching, not prior knowledge removal. If you train people in a classroom and expect them to apply that learning in a very different work environment, you’re creating conditions for the old behaviour to resurface.
Resistance isn’t obstinance; it’s retrieval competition.
When someone keeps defaulting to old ways of working despite training, the most likely explanation is that their old patterns have higher retrieval strength than the new ones you’re trying to establish. They’re not being difficult; their brain is doing what brains do.
Correction requires building, not removing.
The conceptual change literature shows that correcting misconceptions requires building new frameworks robust enough to override intuitive prior beliefs. The misconceptions remain accessible; they become dominated by stronger, more recently practised correct knowledge.
What we should do instead
The goal isn’t to empty cups before filling them. It’s to fill them with something so compelling that older contents are pushed aside. The brain doesn’t have a delete function, but it has excellent mechanisms for learning new responses that outcompete old ones.
This means:
designing training that builds robust new knowledge structures,
providing extensive practice in realistic contexts,
anticipating that old patterns will reassert themselves and building in opportunities for continued reinforcement,
being patient with what looks like regression, understanding that it’s a normal part of how memory works.
It also means being honest with ourselves and our stakeholders about what training can and cannot achieve. We cannot erase what people have learned, but we can help them acquire new knowledge and skills that, with sufficient practice and the right conditions, will come to dominate their behaviour. That’s a meaningful achievement, but it’s a different achievement from the one that unlearning rhetoric promises.
The evidence converges on a solid, if complex, conclusion: the brain manages access to stored information rather than deleting it. What looks like unlearning is the brain prioritising currently relevant information while preserving the capacity to access older learning when circumstances demand it. Understanding this doesn’t make our job easier, but it does make it clearer, and clarity about the challenge we face is the first step toward addressing it effectively.
References
Note: Following feedback in 2025, I have added a clear marker to any references that require payment to access. Whilst I wish I could always offer free access, some sources I rely on can only be found in books, journals, or sites that require subscriptions. I will, as always, do my best to limit these, but at least now you’ll know before clicking.
Vosniadou, S. (2013) ‘Conceptual change in learning and instruction: the framework theory approach’, in Vosniadou, S. (ed.) International Handbook of Research on Conceptual Change. 2nd edn. New York: Routledge, pp. 11-30. (Incomplete source for the entire handbook, but complete for this particular chapter.)
As a reader of the Instructional Design Tips Substack, you can get 25% off a ticket to the very first IDTX Evidence-Informed Practice Conference.
This one-day event is set for the 29th of May 2026 and will be held in Birmingham city centre, UK. The day will see us bring together researchers, scientists, and practitioners to discuss how we utilise the wealth of scientific understanding, research, and evidence to improve workplace training.
To claim your discounted ticket, head over to the IDTX website and use code CPDW25 at checkout.
Nice. Memory doesn't have an eraser, more like a highlighter. New learning competes with old, it doesn't replace it.
The brain also defaults to what is familiar (and assumes it is 'safe'). It is making predictions based on previous patterns or learnings, and will also prioritise what takes the least amount of energy UNLESS specifically trained not to do this in a particular circumstance. Labeling up front what the "outdated" pathway can be helpful. The ideas for embedding this into L&D programmes are great suggestions. This shows up in our daily lives too. Can you (the broad "you", not just you Tom :) ) recognise for yourself where your outdated scripts are showing up? Grocery store, morning routine, parenting, etc. And are there learnings to be updated? I would encourage people to try this for themselves before implementing for an organisation or group. It's how the brain works!