Don't overload working memory

The gist. How many times have you put a diagram next to a paragraph explaining it, and just assumed that combination was helping? Across six experiments, splitting a diagram from its explanatory text only hurt learning when the two had to be mentally combined to make sense. When a diagram already explained itself, adding separate text made scores worse, not better, and simply merging the text onto the diagram didn't fix it.

What the study found

Chandler and Sweller ran six experiments: trade apprentices learning to test electrical wiring, and Year 9 science learners learning how blood flows around the body. When a diagram needed a separate explanation to make sense of it, physically merging the two into one source raised written test scores by roughly two thirds, and the advantage held for 12 weeks.

When the diagram already explained itself without any extra text, the fix was not to merge the text in more neatly. It was to delete it: diagram-only groups outscored both the merged and the separate-text groups, in some cases by roughly double.

What it might mean for your classroom

This is one instance of a broader pattern in cognitive load theory: working memory has limited capacity, and material that forces learners to hold and combine separate sources uses up that capacity before any actual learning can start. The same logic sits behind dual coding, using words and pictures together, but only when the combination is genuinely doing work.

• Move. Cover the text next to a diagram. If a learner can still answer the key question from the diagram alone, the text is redundant.
• Move. If the diagram is unintelligible without the words, put the words directly on the diagram, not in a separate list or paragraph below it.
• Move. If the diagram already explains itself, delete the extra text. Adding it back in, however tidy, tended to lower scores, not raise them.

Treat it as an experiment, not a recipe: try it on one handout this week. Cover the paragraph next to a diagram and check whether learners can still answer your key question from the diagram alone. Merge the words onto the diagram if they cannot, or cut the paragraph if they can. Notice whether learners who usually reread the paragraph twice stop needing to, or answer just as fast once the extra text is gone. If you want a fuller framework for weighing this kind of decision across your own materials, see the guide on reducing cognitive load through lesson design.

How solid is this?

The evidence here is graded Promising: this rates how solid the finding is, not how useful the idea is. The pattern repeated across six experiments and two unrelated subjects, trade apprentices and Year 9 science learners, but it all comes from one research team's 1991 studies, with small groups of 14 to 15 apprentices or 10 to 15 school learners per condition. The authors say so themselves: most of the experiments had no direct measure of cognitive load, so the theory explains the pattern without showing directly what was happening in a learner's head. Half the evidence is adult trade apprentices in workplace training rather than a school classroom, and the school half is Year 9 science learners in a single Australian high school, not a UK sample, and not tested beyond that one year group.

The source

Chandler, P., & Sweller, J. (1991). 'Cognitive Load Theory and the Format of Instruction.' Cognition and Instruction, 8(4), 293-332.

Read the original study