Forecast
Cognitive load theory may be overstating what it explains
The staying power of cognitive load is instructive. It arrived in the 1980s—John Sweller's work on problem-solving—and it felt *right* immediately. We all experience it. Your brain gets full. You can't hold the new information and the old framework in mind at once. The theory formalized that intuition into something testable. Which should've been the start, not the credential.
Here's what troubles me: cognitive load predictions are mostly post-hoc. You redesign a UI to reduce extraneous elements and performance improves—was it cognitive load or just clarity? You present information in a scaffolded way and retention goes up—was it load reduction or was it spacing, or better encoding? The framework absorbs *every* success story about learning or attention without really specifying when it won't apply. That's the telltale sign of folk psychology in respectable clothes. It explains outcomes after they happen, which isn't nothing, but it's not prediction in the sense that matters for science.
The institutional history is telling. You see cognitive load cited everywhere—education, UX, military training—almost as a talisman. It's become the null hypothesis: when something about learning or attention goes wrong, people reach for it. But I've watched projects where teams obsessed over reducing cognitive load in the design phase and still shipped something people couldn't use. The theory didn't catch what was actually throttling performance. Working memory capacity turned out not to be the constraint; the problem was something else—mental model mismatch, poor feedback structure, bad task sequencing.
I'd say it's useful as a *prompt* for attention, not as an explanatory mechanism. When you're designing something educational or interactive, keeping "will this overload someone?" in your head prevents certain obvious mistakes. But it's not a framework that predicts outcomes much better than "make it simpler" does. And that's worth admitting.
1 comment
Log in to comment.
This resonates with what I've seen in classroom research too. There's a study from the late 2010s (Sweller himself co-authored) looking at worked examples—one of CLT's clearest predictions—and the effect sizes were much smaller when researchers actually controlled for prior knowledge and task familiarity. When you account for those, the "load reduction" explanation starts looking thin. The improvement might just be that seeing a solution path first is useful, full stop. Doesn't need the cognitive architecture.
The talisman thing is real. I watched a district sink money into redesigning their curriculum around "reducing extraneous load"—basically cutting out anything not directly tested—and teacher feedback was that students became *more* dependent on scaffolding, not less. They weren't building durable representations; they were just following narrower paths. A colleague argued the district had misapplied the theory, but that's exactly the problem: if your framework can be applied and misapplied in ways that produce opposite outcomes, and both get explained post-hoc, it's doing a lot of work in the interpretive layer rather than in prediction.
Where I'd push back gently on your last point though: it *is* useful as a prompt, but maybe not in the way people think. The value might not be "reduce load" specifically but "be intentional about what you're asking people to attend to." That's good design sense regardless of whether working memory is actually the bottleneck in any given situation.