Cognitive Load During Problem Solving: Effects on Learning

John Sweller is an Australian educational psychologist who spent effectively his whole career at the University of New South Wales in Sydney, where he is now emeritus. He took a doctorate from the University of Adelaide in 1972 on animal discrimination learning, moved to UNSW in 1973, and over the late 1970s and early 1980s shifted from animal paradigms to human problem solving. By 1988 he had been working on the question for several years; this paper is the point at which the strands gathered into a named theory. Cognitive load theory is the work Sweller is known for, and he stayed with it for decades, later folding in an argument from evolutionary psychology about which kinds of knowledge humans acquire easily and which they do not.

The paper appeared in Cognitive Science, the peer-reviewed journal of the Cognitive Science Society — by 1988 a central venue for the field, and notably the journal Allan Collins, a co-author of the cognitive-apprenticeship chapter elsewhere in this set, had helped found. It is a primary source: it reports Sweller's own experiments, several of them with collaborators in the years leading up to it, and it pairs them with a computational model. The experimental work mostly used mathematics and physics problems given to students; the data are the author's own, and the studies he draws together are his and can be traced to their separate papers.

The argument is the founding statement of cognitive load theory. Sweller's claim is that conventional problem solving — in particular the means-ends analysis a novice uses, working backward from the goal and setting subgoals — consumes so much working-memory capacity that little is left over for the thing that actually matters for learning, the building of schemas. A learner can solve a problem and learn almost nothing from having solved it, because the cognitive processes that solving demands and the processes that schema acquisition demands overlap too little. Experts, the paper observes, do not work backward from the goal at all; they recognise a problem type and move forward. The instructional implication, which later cognitive-load research developed at length, is that worked examples and goal-free problems can teach better than conventional problem-solving practice. The position sits in tension with the discovery-oriented and situated approaches of the period, including the cognitive-apprenticeship chapter in this set, and it has a quieter affinity with Bloom's 1984 finding, in "The 2 Sigma Problem", that structured support outperforms leaving learners to work unaided.

Sweller's stake is reputational and professional rather than financial. The paper sold nothing; it is a journal article reporting laboratory studies. It launched a research programme that he led and that carries his name, and that programme has been institutionally successful — cognitive load theory is among the most-cited frameworks in educational psychology and the basis of a large applied literature — so the reputational return on the 1988 paper landing has been considerable. The theory has also drawn methodological challenge over the decades, on the measurement of "load" and on the generality of the worked-example advantage, which a reader meeting the founding paper should weigh against its laboratory origins.