In defence of Cognitive Load Theory
Alfie Kohn misreads both the science and the spirit of Cognitive Load Theory: what he sees as dogma is, in fact, a model of intellectual humility and instructional precision.
I’ve just read Alfie Kohn’s critique of Cognitive Load Theory (CLT) written last year. In it he argues that direct instruction - where teachers explicitly show students what to do and provide ready-made knowledge - is not only pedagogically limited but potentially counterproductive. Drawing on a swathe of research, he contends that inquiry-based, student-centred learning yields stronger results: not only in terms of long-term retention and conceptual understanding but also in motivation, interest, and the development of deeper cognitive capacities. He critiques the recent vogue for CLT as an attempt to bolster a teacher-directed model with shaky theoretical foundations. Kohn claims that CLT is riddled with methodological flaws, assumes an overly simplistic model of memory, ignores the importance of motivation, agency, and social context, and only applies to a narrow range of artificial problems. His overarching claim is that if we care about the kinds of learning that truly matter - critical thinking, transfer, and enduring understanding - then we ought to prioritise rich, exploratory, collaborative approaches, not rigidly sequenced instruction. CLT, he suggests, is a pseudo-scientific smokescreen that props up a regressive educational model.
Kohn is right to highlight the danger of overreliance on narrow, short-term measures of success. Instruction that leads to shallow performance on post-tests should not be mistaken for meaningful learning. He is also absolutely right that students’ motivation, curiosity, and agency are essential to meaningful education. I’d agree that apathy is not a price worth paying for efficiency. His critique of crude, binary comparisons - pure discovery vs explicit instruction - is well taken (except for the fact that he makes his own crude, binary comparison) and his call for nuance in evaluating the complexity of learning processes is entirely fair.
Kohn draws attention to how much of CLT research is grounded in contrived laboratory problems rather than messy classroom realities. This raises important questions about ecological validity. He also draws attention to the difference between learning and performance, a distinction first made prominent by Robert Bjork, which is widely overlooked. His emphasis on the long-term developmental trajectory of learners - including affective, social, and ethical dimensions - is a valuable corrective to overly technical views of teaching.
However, there is a fair bit of his article I want to rebut. Ironically, while decrying caricatures, he conjures several of his own. He portrays CLT as a monolithic and dogmatic framework, yet it is, as Paul Kirschner eloquently argues in fact, a model of epistemic humility, an exemplar of how scientific theories should evolve.
John Sweller’s 2023 article, The Development of Cognitive Load Theory: Replication Crises and Incorporation of Other Theories Can Lead to Theory Expansion is not a defence of CLT’s infallibility, but a celebration of its fallibility as a strength. The very “failures” Kohn gleefully catalogues - modality reversals, elusive effects, contradictory results - are not damning. They are catalytic. Each empirical hiccup has ended up refining rather than collapsing the theory. CLT expanded to account for new variables - element interactivity, expertise reversal, the distinction between intrinsic and extraneous load - not because it was ideologically rigid, but because it took its own limits seriously.
Where Kohn sees a pseudo-theory bloated by retrofitted constructs, Sweller sees a model in recursive repair. CLT has absorbed insights from memory research, developmental psychology, and even evolutionary theory. It doesn’t pretend to predict everything. But it does offer a principled, falsifiable framework grounded in the architecture of cognition, which is more than can be said for many pedagogical manifestos.
Let’s take one such refinement: element interactivity. Kohn derides CLT’s lack of nuance, yet this concept is the very opposite. It recognises that what constitutes complexity depends on what the learner already knows. For novices, tasks with many interacting elements (such as algebraic problem solving) impose overwhelming load. For experts, those same elements become single “chunks” retrievable from long-term memory. This matters because it reveals why instructional approaches must be stage-dependent. What works for experts can bewilder beginners.
Recent neuroimaging evidence further strengthens the case. Erol Ozcelik’s 2025 fMRI study on graph comprehension1 directly demonstrated that higher cognitive load -induced through split-attention designs - corresponded to increased activation in the brain’s frontoparietal and multiple-demand networks. These are the same domain-general systems responsible for juggling working memory, attention, and cognitive control. This matters because it shows cognitive load is not just a theoretical construct but a biological reality, measurable in neural terms.
In contrast to Kohn’s claim that CLT relies on untestable constructs, Ozcelik’s study shows that excessive cognitive load ‘lights up’ domain-general networks that juggle attention, inhibition, and working memory, just as the theory predicts. The research underscores that when instructional design imposes unnecessary demands — as with split-attention formats — learners’ cognitive architecture is overwhelmed, impairing performance. Far from being a dogma in search of evidence, CLT now draws strength from converging behavioural, physiological, and neuroimaging data.
Kohn also ignores the biological turn in CLT, perhaps its most radical implication. Drawing on David Geary, Sweller distinguishes between biologically primary knowledge (language, face recognition, social cues) and biologically secondary knowledge (reading, mathematics, scientific reasoning). We are evolutionarily primed to learn the former through immersion and exploration. The latter, however - the stuff of school - is not naturally acquired. It must be explicitly taught.
This evolutionary lens resolves the romantic notion that all learning should feel “natural.” Reading isn’t natural, neither is writing an analytical essay or solving a quadratic equation. They require instructional design, because they demand we process unfamiliar, high-element information under severe cognitive constraints. This is, as I argued here, the very reason schools exist.
Kohn might protest that explicit instruction stifles curiosity but he overlooks the evidence that well-sequenced, explicit instruction enables inquiry and epistemic curiosity by furnishing students with the very schemas and concepts they need to explore effectively. Struggle only becomes productive once students have enough background knowledge to make the effort meaningful. Without that foundation, we create what Sweller calls undesirable difficulties: situations in which problem solving interferes with learning.
Most egregiously, Kohn’s critique conflates direct instruction with mindless “chalk and talk.” He neglects the work of Rosenshine, or the many contemporary teachers who use explicit instruction as a launchpad for thinking. Direct instruction is the most active form of teaching that requires all students are actively involved throughout lessons. It means starting with clarity, modelling complex processes, constantly checking understanding, and then gradually releasing responsibility.
Lastly, Kohn implies that progressive education is simply more humane. But here he commits a subtler error: he mistakes affective preference for instructional effectiveness. Maybe collaborative inquiry can be delightful, maybe there are students who prefer it but these are likely to be those who are already most advantaged. If we care about equity, about ensuring that disadvantaged children -those without the luxury of home capital - grasp the curriculum, then we must care about what works. And what works best, at scale, to teach biologically secondary knowledge to novices is explicit instruction, informed by CLT and tempered by teacher judgment.
For me, equity is the most important consideration when designing instructional sequences. The approaches advocated by Kohn are only likely to be effective for the most privileged and are further disadvantage the most disadvantaged. It’s hard to argue that there’s anything human about widening the advantage gap.
In conclusion, Kohn’s essay is engaging, but it commits a series of category errors. It treats instructional preference as moral certainty. It dismisses CLT because it evolves. It caricatures direct instruction while romanticising inquiry. And it forgets the central insight of CLT: that learning is not about what feels good in the moment, but about what endures.
We should not accept any theory uncritically. But nor should we reject it because it makes us uncomfortable. As Kirschner and Sweller show, CLT is not a dogma but a living theory - evolving, integrating, refining - just as real learning should.
Ozcelik, E. (2025). The neural correlates of cognitive load in learning: An fMRI study on graph comprehension. Learning and Instruction, 99, 102175. Unfortunately, this is currently behind a paywall.
Always appreciate your writing, David. I also appreciate critical debate for the ways it crystallizes my own perspective.
I was a student of Kohn’s perspective on student engagement back in the early 2000s, and I would say the legacy of that still holds in my perspective on teaching and learning - that said, I ALSO believe explicit instruction (and other direct instruction moves) gives students the best shot at the taste of success they need to develop the intrinsic motivation that will sustain their growth in learning.
I will say I do get very frustrated with the “methodical weakness” argument. In my humblest opinion, all of education research has methodical weakness when it comes to applying theories in practice. Theories are underpinnings, and methods are potential actions. “Hard science” requires controlling for variables that make outcomes too decontextualized. All
of it - theory, method, education science - requires contextualization - not blind allegiance.
There is also seemingly as issue with semantics. I’m cautious about applying the term “direct instruction” because of the conjurations that Kohn seems to make about it. He goes with the “passive vessels” argument against the “banking model” of education - assuming that direct instruction means lecture. I would agree with him, and Friere, and so on if I believed direct instruction was lecture. But in my work (in the US, by the way, where standards and high stakes tests are God - and where funding for vulnerable populations continues to be slashed), I do not see the same binary that Kohn does. The kind of instruction that I espouse and peddle to teachers and schools is (in short) clear, structured, and gradually released to student independence - at which point all kinds of inquiry and discovery can happen with the knowledge necessary to do so. And I insist that teachers adjust the “instructional method” to fit their content, student needs, and other contextual factors - and I assist them in doing so. Like you attest, their judgment and ownership is critical.
Perhaps it’s because I left academia to return to the “trenches” of public schools (much to the dismay and pity of many of my IHE colleagues)- and I see and work in schools and with students in places where 100% economically disadvantaged is but one feature - but I must admit I grow weary with the debate for sake of it. Inquiry learning as constructivist method must be constructed from something that many students simply do not have - yet. That’s a reality.
And while we sit around arguing, Rome is burning.
Who is the audience Kohn is after? Who is the audience any of us is after? Who are we trying to convince? One another? How does this translate to meaningful change at the student level? And who is responsible for that? This is my struggle. For all the good work of scholars in education, I’m still most continually struck by how it makes its way to the ones who actually do the work we write about.
Thanks, as always, for a thought-provoking piece.
There’s an emerging body of evidence to show that load reduction instruction is also good for student motivation and wellbeing. The evidence isn’t just in lab style experiments but correlational, large scale, process product etc. I mean how much more evidence do these people need?