Why the Right Amount of Help Keeps Changing: Cognitive Load

What the theory establishes

Human working memory is narrow, and a vast body of experimental work has demonstrated how narrow. The practical consequence is that teaching succeeds or fails largely on how much it asks a learner to hold at once while something new is being assembled in long-term memory. Learning, in this account, is the development and automation of cognitive schemas: organised patterns that let a mind treat what were once many separate items as a single one.

Cognitive load theory sorts the demand on a learner into two kinds. Intrinsic load is imposed by the material’s own difficulty, measured against what the learner already knows. Extraneous load is imposed by the way the material is presented, and it contributes nothing to learning. A 2017 review in the Journal of Taibah University Medical Sciences, by Jimmie Leppink of the School of Health Professions Education at Maastricht University, sets out what the theory has established and where it has stalled. The paper reports no new data, and the empirical work it draws on was conducted with medical, pharmacy and undergraduate students, all of them adults.

The documented sources of wasted demand are unglamorous. Attention divided between a diagram and a caption sitting elsewhere on the page, or between a demonstration and an explanation that arrives minutes later, spends capacity on reassembly. Information delivered in words when it ought to have been shown does the same. So do confusing instructions from a supervisor, and so do the learner’s own thoughts and feelings while the work is going on.

The best documented remedy is the worked example. A beginner facing an unfamiliar procedure spends much of their capacity searching for a method they do not yet have, and studying a completed, successful solution first removes that search. The effect is robust in early stage learners, and it is the basis of a great deal of sound teaching practice.

As learners become more proficient, the benefit of the worked example fades and eventually reverses, because support supplied where it is no longer needed becomes one more thing to process. Prior knowledge is doing the work in both directions: the more developed and automated a learner’s schemas, the lower the intrinsic load the same material imposes, and the less room there is for help that once paid for itself.

What this means for you and your child

Scaffolding is calibrated to a stage, and stages move. The step-by-step model that carried a child through long division in September can be the thing slowing them in November, and the change has nothing to do with the child’s motivation or the quality of the model. A drop in fluency after support was added is a reason to review the support before adding more of it.

Whenever the material is genuinely new again, the reverse holds. A child who has outgrown worked examples in arithmetic will need them on the first day of algebra, and withdrawing them on the grounds of maturity restores the search cost the examples existed to remove. What matters is the stage the child has reached in that specific material, this term.

Whether a given demand counts as waste depends entirely on the learning goal. A ringing phone and colleagues passing through the room are extraneous load when the goal is clinical reasoning, and intrinsic load when the goal is learning to work amid interruption. This is a review of adult professional education, and no study of children is discussed in it, so its application to a nine year old is an inference. Applied to a child, the same test still decides whether background noise or an unfamiliar worksheet layout is a problem worth engineering away, and the answer changes with what the lesson is actually for.

The design principles are well supported; the measurement of load is not, and the author is direct about the gap. Attempts to capture the different types of load by questionnaire have not produced the correlations with learning that the theory predicts. Single-item ratings are unreliable and cannot distinguish one type of load from another. Ratings of mental effort appear to reflect intrinsic load almost exclusively, and averages taken repeatedly during a task run lower than a single rating taken at the end, so even the number’s size depends on when it was requested. Brain imaging and eye-tracking may, under certain assumptions, index overall load, and no measure of the separate types has yet emerged from them.

So when a school or a programme offers you a figure for your child’s cognitive load, it is claiming a precision that the theory’s own researchers say the field does not yet possess.