When does making learning harder backfire?

When the learner is a beginner and the difficulty is more than they can handle. A desirable difficulty is a condition that takes more effort but improves long-term performance, but only if the task can be accomplished; too difficult a task may dissuade the learner and prevent full processing. And what counts as too difficult depends on what you already know.

The idea comes from cognitive load theory, which John Sweller developed in the late 1980s from a study of problem solving. Its starting point is that working memory is extremely limited in capacity and duration. Beginners lack ready-made knowledge structures, so if instruction gives them no guidance they often fall back on inefficient problem-solving strategies that overwhelm working memory. That is why worked examples, step-by-step demonstrations of how to solve a problem, are so effective for them.

Then comes the twist, called the expertise reversal effect: techniques that work for one level of prior knowledge reverse for another. For learners with more knowledge, reduced guidance often gives better performance than well-guided instruction, because extra guidance means processing redundant information and more load. Low-knowledge learners benefit more from structured worked examples than from solving problems alone, but as knowledge increases, open-ended problem solving becomes the more effective activity.

One studied response is to fade the help. Worked examples can have their solution steps successively removed, and fading that adapts to each learner's demonstrated understanding gave even better knowledge acquisition than fixed fading.

The field is still arguing about it. Some researchers suggested the testing effect might disappear for more complex material, though later studies showed that complex learning benefits from retrieval practice, and a 2024 paper warned that cognitive science's recommendations cannot be one-size-fits-all.