Desirable difficulties
Also known as productive struggle, storage strength vs retrieval strength
Some kinds of difficulty during study slow you down now and leave more behind later. Others are just difficulty.
The claim Conditions that impair performance during learning — spacing, interleaving, testing, varied practice, generating rather than reading — often improve long-term retention and transfer.
Good evidence The individual difficulties are well replicated; what remains genuinely unsettled is how to tell in advance which difficulties are the desirable kind.
Robert Bjork's phrase is the organising idea behind most of this section. Spacing, interleaving, testing, varying the conditions of practice, generating an answer instead of reading one — five separate literatures, one shared signature. Each makes performance worse while you are learning and better when you are tested later. Bjork's argument is that this is not a coincidence but the same effect seen five times: retrieval is what builds memory, and anything that makes retrieval harder while still succeeding builds more of it.
The distinction that does the work is between storage strength — how well something is learned — and retrieval strength — how available it is right now. Massed rereading raises retrieval strength and leaves storage strength almost untouched. A difficult, delayed, successful recall raises storage strength precisely because retrieval strength had fallen. This is why forgetting a little before you review is not a failure of the schedule; it is the schedule working.
The word doing the most work in the phrase is 'desirable'. Not all difficulty helps. A difficulty is desirable when the effort goes into the thing you are learning; it is merely difficult when the effort goes anywhere else — a bad interface, an unreadable font, material three levels above you, a distraction you have to keep suppressing. Cognitive load theory is the systematic account of that second category, and the two frameworks are describing opposite ends of one dimension rather than disagreeing.
There is a real practical problem here that the popular accounts skate over. The theory tells you that useful difficulty and useless difficulty feel the same from the inside. Both are unpleasant, both lower present performance, and the only reliable way to tell them apart is a delayed test — which means you find out weeks after you needed the answer.
The honest working rule most researchers offer: a difficulty is likely to be desirable if you can still succeed at the task, slowly, using what you are trying to learn. If you are failing outright, or succeeding by using something else entirely, you have found the other kind.
What to do differently
- Prefer generating over recognising — write the answer before you look at the options, say the sentence before you read it.
- Let a little forgetting happen before you review. The effortful recall is the mechanism, not the side effect.
- Vary the conditions of practice: different rooms, different orders, different examples of the same idea.
- Check that the difficulty is in the material. Fighting a tool, a font or an unusable app is load, not learning.
- If you are failing outright rather than succeeding slowly, step back a level. Desirable difficulty requires that you can still get there.
Limits and trade-offs
- There is no reliable in-the-moment test for whether a difficulty is the good kind, which is a serious gap in a theory offered as practical advice.
- For a learner with weak prior knowledge, most added difficulty is the wrong kind — the worked-example literature finds the opposite prescription for beginners.
- It is easy to abuse as a justification for bad teaching or bad materials: any complaint can be reframed as productive struggle.
- The benefit is measured on delayed tests. If your actual goal is short-term performance, the trade genuinely goes the other way.
What this means for a language
This is the cleanest explanation of why comfortable study so often leads nowhere. Rewatching an episode you already understand, rereading a text you have already parsed, doing exercises on the tense you are best at — all fluent, all nearly free of gain. It also explains why the first weeks of unsupported listening feel like drowning and are nonetheless the thing that works. The counterweight matters just as much: input far above your level is not a desirable difficulty, it is noise, and hours of it produce very little. The target is material you can just barely follow.
The research
- Bjork, Dunlosky & Kornell (2013), Annual Review of Psychology
Sets out the storage/retrieval-strength distinction and reviews the difficulties — spacing, interleaving, testing, generation — that share the signature.
- Rohrer, Dedrick, Hartwig & Cheung (2020), Journal of Educational Psychology
A worked case of the signature: interleaving lowered accuracy during practice and raised it a month later, 61% to 38%.
Related
Read next
-
The review that explains why good technique is not enough: learners systematically misjudge their own learning, and the methods that feel best while studying are usually the ones that leave least behind.
-
The standard first book. Two of the three authors are the cognitive psychologists behind much of the retrieval-practice literature, so the claims are first-hand rather than second-hand. Heavy on anecdote between the findings, which some readers find padding.
-
Popular-press, and it argues a thesis rather than surveying a field — but its chapters on interleaving, desirable difficulties and the limits of early specialisation are the most readable version of that literature anywhere.
-
The single most useful document here. Ten study techniques rated for utility across materials, learners, tasks and settings. Only practice testing and distributed practice earn 'high'; highlighting, rereading and summarising come out low.