Cognitive load theory: designing for the bottleneck

Evidence grade: STRONG for the core instructional effects (worked examples, split-attention, redundancy, expertise reversal), replicated across decades and domains. Some internal theoretical accounting (how to measure the load types independently) remains debated among researchers — the design prescriptions stand on the experimental effects, not the bookkeeping.

Every idea a child learns passes through working memory — a bottleneck famously capacious enough for only a handful of novel elements at once (Miller's "seven, plus or minus two," 1956; closer to three or four for genuinely novel material in later estimates). Long-term memory, by contrast, is effectively unbounded, and what's stored there — organized as schemas — passes through the bottleneck as single chunks. Cognitive load theory (Sweller, 1988; Sweller, van Merriënboer & Paas, 1998, 2019) is the systematic working-out of one design question: is the instruction spending the bottleneck on learning, or wasting it?

The theory's useful vocabulary: intrinsic load (the task's inherent complexity — how many elements interact), extraneous load (imposed by presentation, not content — the waste), and the processing that builds schemas (historically "germane"). The designer's job: manage intrinsic, minimize extraneous, leave room for the building.

The effects that matter for teaching

The worked-example effect (Sweller & Cooper, 1985, and a large literature since): novices learn more from studying worked solutions than from solving equivalent problems unaided. Problem-solving search consumes the bottleneck; a worked example spends the same capacity on the solution's structure. The practical form is the fading sequence: worked example → completion problems (partial solutions) → full problems.

The expertise reversal effect (Kalyuga et al., 2003): the worked-example advantage reverses as knowledge grows — for learners with established schemas, studying examples becomes redundant and solving becomes superior. No method is best independent of the learner's current knowledge; sequencing is everything.

The split-attention effect: when understanding requires integrating separated sources (diagram here, explanatory text there), the integration itself burns working memory. Physically integrate them — labels on the diagram, instructions beside the thing they describe.

The redundancy effect: presenting the same information in duplicate simultaneous forms (reading text aloud verbatim while displaying it; decorating a clear diagram with restating text) hurts — processing the duplicate is pure extraneous load. More supports are not automatically kinder.

The guidance argument: Kirschner, Sweller and Clark's much-cited 2006 review argues from the same architecture that minimally-guided discovery fails novices — they search where they should be building. Read carefully, this is not an argument against all inquiry; it's an argument about novices, fully consistent with reducing guidance as expertise grows (the reversal effect again).

What the evidence doesn't say

It doesn't say "make everything easy." Desirable difficulties and CLT divide the territory: CLT eliminates difficulties that are presentational waste; the desirable ones (retrieval, spacing) engage learning processes. A clean instruction card followed by an answer-free retrieval demand is both theories at once.
It doesn't license dull minimalism. Motivation matters; the theory only insists that decorations not sit between the learner and the structure being learned.
It doesn't fix the load numbers precisely — treat "3–4 novel elements" as an engineering constant with error bars, not scripture.

In the classroom

For new material: worked examples first, then fade. "Watch one, complete one, do one" is CLT in a sentence.
Put words next to their referents — instructions inside the interface being taught, labels on the diagram, the hint beside the field it concerns.
Kill redundancy: don't read slides aloud verbatim; don't caption what the picture already says (presentation literacy is this effect taught to children).
Sequence by element interactivity: teach interacting elements (formula + cell references + fill-down) in cumulative steps small enough that each fits the bottleneck with room to spare.
Automatize the tools: fluent typing and mouse use are schema-side capacity — every skill left unautomatized is a permanent tax on the bottleneck during everything else (why typing fluency matters).

How Wiz Kids applies this

Task instructions live inside the simulated interface, next to what they describe (split-attention); teaching tasks demonstrate and scaffold while review tasks strip guidance (fading + expertise reversal, mechanized); lessons introduce one or two interacting elements at a time along a prerequisite graph that keeps intrinsic load bounded; and the whole typing thread exists because automatized transcription frees the bottleneck for everything downstream.

References

Miller, G. A. (1956). The magical number seven, plus or minus two. Psychological Review, 63(2).
Sweller, J. (1988). Cognitive load during problem solving: Effects on learning. Cognitive Science, 12(2).
Sweller, J., & Cooper, G. A. (1985). The use of worked examples as a substitute for problem solving in learning algebra. Cognition and Instruction, 2(1).
Sweller, J., van Merriënboer, J. J. G., & Paas, F. (1998; updated 2019). Cognitive architecture and instructional design. Educational Psychology Review, 10(3); 31(2).
Kalyuga, S., Ayres, P., Chandler, P., & Sweller, J. (2003). The expertise reversal effect. Educational Psychologist, 38(1).
Kirschner, P. A., Sweller, J., & Clark, R. E. (2006). Why minimal guidance during instruction does not work. Educational Psychologist, 41(2).

© Glu IO Pty. Ltd. — Wiz Kids (wiz.kids). Link freely; republication requires permission — see terms. Found an error in our reading of the research? We correct fast: tell any teacher piloting Wiz Kids.