Desirable difficulty: why the struggle is the point

Spacing your reviews out feels worse than cramming. Testing yourself feels worse than rereading. Mixing your cards feels worse than sorting them by theme. And in study after study, the worse-feeling option wins the delayed test.

Once is a quirk. Three times is a pattern, and the pattern has a name. Robert Bjork called these desirable difficulties in 1994, and the idea has held up so well that most of modern learning science reads like footnotes to it.

01 / 06

Effort buys retention; smoothness buys performance that evaporates

Here's the claim. Conditions that slow down apparent learning and introduce effort often produce better retention and transfer (using what you learned in a new situation). Conditions that make learning feel smooth often produce performance that evaporates.

The word doing the work is desirable. Each of these hardships forces the memory system to do something it would otherwise skip: reconstruct instead of recognize, discriminate instead of coast, rebuild instead of glance.

Bjork's 1994 chapter was written for training designers, and its charge sheet still reads fresh. We build practice to maximize how learners look and feel during the session, then act surprised when the gains evaporate.

Thirty years later most study apps still make the same trade, because smooth sells. It just doesn't store.

02 / 06

Accessibility fades; depth, once built, essentially never does

Bjork's explanation runs on a distinction his lab formalized as the new theory of disuse (Bjork & Bjork, 1992). Every memory has two separate strengths:

• Retrieval strength: how accessible it is right now. High after you've just seen it, decays fast.
• Storage strength: how deeply it's woven in. Builds slowly, essentially never decreases.

The counterintuitive part, and the engine of everything: the lower the retrieval strength when you successfully recall something, the bigger the gain in storage strength. A word retrieved while it's nearly gone gets welded in. A word retrieved 10 seconds after you saw it gains close to nothing.

A recall made at the edge of forgetting is worth far more than a fresh one Retrieval strength is how accessible a word is right now; storage strength is how deeply it is woven in. The lower the first is when you successfully recall, the more the second gains.

• Recalled while fresh, seconds after you saw it: close to no gain in storage.
• Recalled at the edge, when it had nearly faded: the word gets welded in.
• Left to right: how far the word had faded before you pulled it back.

Framework: Bjork & Bjork 1992, the new theory of disuse. Shape is illustrative. The theory specifies the direction, not a measured curve.

Read your study habits through the two-strengths lens and everything inverts. The cram feels great because retrieval strength is pegged at maximum the whole time, which is exactly why storage strength never moves. The spaced review feels shaky because retrieval strength has sagged, which is exactly why the recall pays out.

03 / 06

The learner who felt worse wins, and never believes it

Make it concrete. Two learners, same 20 free minutes on a Tuesday, same word list from last week's lesson.

Learner A opens the list and reads it through 4 times. Every word is right there, recognition fires instantly, and the session feels like polishing silver. Twenty minutes of 100% success. She closes the laptop feeling fluent.

Learner B does 20 minutes of due cards. Half the words have gone fuzzy since last week, so the session is full of pauses, two outright blanks, and one word she gets only after staring at the ceiling. Maybe 80% success, none of it smooth. She closes the laptop feeling like Spanish is fighting her.

Test Learner A and Learner B in 2 weeks and B wins, clearly. Every one of her ugly pauses was a low-retrieval-strength recall, the exact event that welds storage strength. A's silky rereads were high-retrieval-strength passes, pleasant and nearly worthless. A spent 20 minutes feeling good. B spent 20 minutes learning.

The kicker: ask both learners afterward who had the better session, and A says A, with total sincerity. Her evening felt more competent. This mismatch between judgment and reality has its own literature (start with how to grade yourself honestly), and it's why you can't navigate by feel here.

04 / 06

Four separate effects are one lever: how faded the word is when you test it

Bjork's framework earns its keep by predicting a whole family of effects that were discovered separately:

• Spacing. Waiting lets retrieval strength drop, so each review lands harder. 317 experiments agree (Cepeda et al., 2006).
• Testing. Retrieval is effortful reconstruction; rereading is recognition, which costs nothing and builds about as much. The effortful one roughly doubled week-later recall in Karpicke and Roediger's Swahili study.
• Interleaving. Mixing categories keeps you off balance and forces discrimination. Doubled delayed math scores in Rohrer's classroom work.
• Generation. Producing an answer, even a wrong one, beats being handed it (Slamecka & Graf, 1978; Kornell, Hays & Bjork, 2009 showed even failed attempts help).

Four literatures, one shape: add the right kind of effort, lose a little today, keep a lot next month.

The overlap has a cause. Each technique is a different lever on the same variable, the retrieval strength of the memory at the moment you exercise it. Lower it (with time, with mixing, with a blank instead of an answer) and the exercise pays more.

05 / 06

When does a difficulty stop helping you learn?

A difficulty stops helping once you can't get past it, and that warning label gets dropped in the retelling. Bjork's own phrasing: a difficulty is desirable only if the learner can overcome it. Struggle that ends in reconstruction builds memory. Struggle that ends in nothing builds frustration.

Some difficulties are plainly undesirable. Cards you fail 5 times running (the interval was too long; the scheduler should pull them closer). Audio too fast to parse at all (that's noise, and you can't acquire from noise). Study sessions so long that fatigue, a completely undesirable difficulty, swamps everything.

Keeping difficulty on the desirable side is the tightrope a good scheduler walks for you. Verbamor's target is a roughly 90% success rate: hard enough that most retrievals happen near the edge, gentle enough that you keep winning. Difficulty, dosed.

06 / 06

Fluency during study is a broken gauge, so read a delayed one

The last problem is emotional, and I don't want to wave it away. Every desirable difficulty produces the subjective feeling of being worse at Spanish than you were during the smooth, useless version of studying. Learners quit over that feeling. Kornell and Bjork's painting participants sat in front of their own superior results and still believed in blocking (studying one painter at a time).

Two reframes help, and both happen to be true:

• The strain of a slow retrieval is the sensation of storage strength increasing. That specific discomfort is the product working. You're feeling a weld.
• Fluency during study predicts almost nothing. So judge your progress by delayed evidence instead: last month's words showing up, unprompted, in this week's tutor session.

The delayed evidence is also checkable, which helps on bad days. Your review history shows intervals stretching month over month: a word that once needed reviewing every 2 days now waits 6 weeks and still lands. That chart is storage strength made visible, and it only ever grew during the sessions that felt like sand.

One more habit worth building: when a session runs hard, resist the instinct to "fix" it by studying longer that night. The difficulty was the dose, and the dose was delivered.

More smooth reps after the struggle mostly buy back the good feeling. They cost you your evening (and your sleep, which has its own job to do with what you just learned).

You'll never learn to love the wobble, exactly. But you can learn to read it correctly, and learners who do stop mistaking smooth for strong. The struggle was the learning all along.

---

Sources