Desirable Difficulties: S-Tier Behavioral Designer’s Guide

Tell a learning scientist you have made your training “effortless,” and you have just told them you have made it forgettable. To them, the two words point at the same outcome. Every minute a new skill feels easy to acquire is, very often, a minute it is quietly failing to lodge in long-term memory. That claim is not a hot take. It rests on forty years of controlled experiments, and it is almost completely ignored by the industry that needs it most.

Robert Bjork, the UCLA memory researcher who named the idea, calls these conditions desirable difficulties: ways of arranging practice that slow you down, make you stumble, and feel like worse progress in the moment, yet produce dramatically more durable and transferable learning later. Space your study sessions instead of cramming. Mix problem types instead of drilling one. Test yourself instead of rereading. Each one hurts your performance today and helps your retention next month.

Here is why this should keep every product designer, educator, and marketer awake. The entire modern doctrine of user experience is “remove friction.” Make it smooth. Make it effortless. And the science of how humans actually retain anything says that for whatever you want a person to learn, remember, or master, smoothness is the enemy. This guide will show you what desirable difficulties are, why they work, where they break, and the one distinction that separates the friction you must remove from the friction that is doing all the work.

Speed Run Notes

• Desirable difficulties (Robert Bjork, 1994): practice conditions that feel harder and slow you down now but produce far more durable, transferable learning later. The struggle is not the cost of learning; it is the learning.
• The load-bearing distinction is learning versus performance. Performance is what you can see during training (fast, fluent, looks like progress). Learning is the durable change that shows up on a delayed test. The conditions that boost one often wreck the other.
• The core set: spacing beats massing, interleaving beats blocking, retrieval practice beats rereading, generating beats being shown. All four feel worse in the moment and win on the delayed test.
• Difficulty is only desirable if the learner can overcome it. Past that line it becomes an undesirable difficulty: load that blocks learning instead of building it. The font-disfluency studies are the cautionary tale here, with mixed and failed replications.
• For designers: your dashboard measures performance during the session. The thing your business runs on (retention, mastery, durable habit) is learning, and learning is invisible on the acquisition dashboard. You can win every in-session metric and lose the only one that matters after.
• The Octalysis link: desirable difficulty is the raw material of Core Drive 2 (Development & Accomplishment). Remove the difficulty and the accomplishment goes hollow. Frictionless UX does not just hurt memory; it starves your most reliable motivation driver.

Author Credibility: Yu-kai Chou

Yu-kai Chou created the Octalysis Framework after studying gamification since 2003 — years before the term entered mainstream vocabulary. As a Human-Systems Architect & Behavioral Designer, his framework has been applied by LEGO, Microsoft, Porsche, Coca-Cola, Salesforce, and MrBeast, impacting over 1.5 Billion Users.

Chou has taught the Octalysis methodology at Harvard, Stanford, Yale, Tesla, Google, BCG, and IDEO.

His work has been cited by Harvard, Stanford, MIT, Forbes, Wall Street Journal, Wired, US Department of Energy, NIST, NSF, NCBI, US Department of Education, ClinicalTrials.gov, and Google Scholar — with 3,700+ more academic publications. Explore his books here.

What Are Desirable Difficulties?

A desirable difficulty is a condition of practice that introduces effort, slows acquisition, and lowers your visible performance during learning, while increasing how much you actually retain and how well you can apply it later. Bjork introduced the term in a 1994 chapter on the training of human beings, and the word that carries the weight is “desirable.” Most difficulty is just difficulty: it makes things harder and helps nothing. A desirable difficulty is the specific kind that the brain converts into durable memory precisely because you had to work for it.

The classic demonstration is almost insultingly simple. Give one group of students a list to study four times in a row. Give another group the same list, but space the four study sessions across days. Test both groups ten minutes later and the crammers look like geniuses. Test both groups a week later and the spacers crush them. Same material, same total study time, opposite outcomes. The only difference is that spacing made each session feel harder, because you had partly forgotten the material and had to fight to bring it back.

That fight is the mechanism. When retrieval is easy, almost nothing happens to memory. When retrieval is effortful but successful, the memory is strengthened. Difficulty, in the right dose, is not interference with learning. It is the trigger for it. This is the idea that flips the intuition of nearly every learner and every designer: the discomfort you have been trying to engineer away is the thing you should have been protecting.

The catch, and Bjork is explicit about it, is the dosage. A difficulty is desirable only if the learner has the background and resources to overcome it. Hand a beginner a problem they cannot possibly solve and you have not created a desirable difficulty. You have created frustration, a wall, a reason to quit. The art is in finding the band of difficulty that is high enough to demand real effort and low enough to be conquered. Everything useful about this theory lives inside that band.

Learning Is Not Performance

To use desirable difficulties without hurting yourself, you have to internalize one distinction, and it is the most underrated idea in all of applied psychology. Nicholas Soderstrom and Robert Bjork laid it out in a 2015 review that every person who builds a teaching system should read: learning is not the same thing as performance, and performance is an unreliable guide to learning.

Performance is what you can observe and measure right now, during or immediately after training. It is the quiz score at the end of the lesson, the completion rate of the tutorial, the speed at which someone finishes the task today. Learning is the relatively permanent change in knowledge or skill that supports retention and transfer over the long term. You cannot see learning directly. You can only infer it later, from a delayed test or a transfer to a new situation.

The reason this matters so much is that the two can move in opposite directions. Soderstrom and Bjork document case after case where a manipulation boosts performance during acquisition while reducing long-term learning, and others where a manipulation depresses performance during acquisition while increasing it. Massed practice, immediate feedback after every step, and keeping conditions constant all make you look better now and learn less. Spacing, delayed feedback, and varied conditions all make you look worse now and learn more.

Sit with the consequence. If you judge a learning experience by how it performs in the moment, you will systematically choose the version that teaches less. You will reward fluency and punish the very friction that builds durability. This is not a rare edge case. It is the default failure mode of anyone who optimizes for what they can measure during the session, which is to say almost every team that has ever run an experiment on an onboarding flow. We will come back to that, because it is the whole game.

One more layer makes it worse. Learners themselves cannot tell the difference. People reliably rate the easy, fluent condition as more effective even after their own delayed test scores prove the opposite. The feeling of learning and actual learning are loosely coupled at best, and the feeling tracks fluency, not durability. So both the designer and the user are pulled toward the same wrong answer by the same illusion.

The Five Difficulties Worth Wanting

Researchers have catalogued a handful of difficulties that reliably pay off. Five of them are well-supported enough that you can build a system around them. Each one shares the same fingerprint: it lowers performance during practice and raises it on a delayed test.

Spacing Beats Massing

Distribute practice across time instead of clustering it. The spacing effect is one of the most replicated findings in the history of psychology, confirmed in a large quantitative review of distributed-practice studies. Cramming produces a strong but fragile spike. Spacing produces a slower build that lasts. The discomfort of spacing is real: when you return after a gap, you have forgotten some of it, and pulling it back feels like failure. That partial forgetting is exactly what makes the retrieval valuable.

Interleaving Beats Blocking

Mix different types of problems or examples instead of grouping them. Nate Kornell and Robert Bjork ran a study where people learned to recognize the styles of different painters. One group saw each artist’s paintings in a block; another saw them shuffled together. Blocking felt far better and the learners said so. But the interleaved group was much better at identifying the artist of new paintings they had never seen. Mixing forces the brain to discriminate between categories rather than lulling it with sameness, which is why interleaving improves the ability to tell things apart and apply them in new cases.

Retrieval Practice Beats Rereading

Trying to recall something strengthens it more than reviewing it. The testing effect, demonstrated cleanly by Henry Roediger and Jeffrey Karpicke, shows that students who are tested on material retain far more on a delayed exam than students who simply restudy it for the same amount of time. Pulling a memory out of your head is an act of construction that reinforces the path back to it. Rereading feels productive because the words look familiar, but recognition is not recall, and familiarity is the fluency trap wearing a lab coat.

Generation Beats Being Shown

Producing an answer yourself, even a wrong one, beats being handed the answer. When you struggle to generate a solution before seeing the correct one, you build a richer memory than if you had read the solution directly. The same logic explains why reduced or delayed feedback often outperforms instant feedback. Immediate correction after every step removes the productive interval in which the learner would have noticed and repaired the error on their own. The struggle to self-correct is where the skill forms.

Varying the Conditions of Practice

Practice the same skill under changing conditions instead of identical ones. Richard Schmidt and Robert Bjork showed in a 1992 synthesis of motor and verbal learning that varying the parameters of practice (the distances, the contexts, the surface features) depresses performance during training and improves both retention and the ability to generalize to a situation you never rehearsed. A child who practices basketball free throws from a single fixed spot looks sharper in practice than one who shifts distances every few shots, yet the one who varied does better in a real game from a spot neither of them drilled. Constant conditions build a memory that is tightly bound to those exact conditions. Varied conditions build a flexible representation that transfers. The cost, again, is a worse-looking practice session and a better real-world result.

Notice what these five have in common beyond the delayed-test payoff. Every one of them is something a well-meaning designer would strip out in the name of a smoother experience. Spacing looks like a slower funnel. Interleaving looks like a confusing curriculum. Retrieval looks like making the user do work the app could do for them. Generation looks like withholding help. Varied practice looks like an inconsistent, less polished experience. The instinct to remove friction targets all five precisely.

Why It Works: Storage Strength vs Retrieval Strength

Bjork’s explanation for all of this is a model he and Elizabeth Bjork call the New Theory of Disuse, and it rests on splitting memory strength into two independent quantities. Every memory has a storage strength, which is how deeply it is embedded and never decreases, and a retrieval strength, which is how accessible it is right now and fluctuates constantly with use and time.

When you cram, you spike retrieval strength. The material is highly accessible immediately, which is why crammers ace a test ten minutes later. But cramming adds very little storage strength, so retrieval strength decays fast and a week later the memory is gone. Spacing does the opposite. Each effortful retrieval after a gap adds storage strength, the kind that lasts.

The counterintuitive engine of the whole theory is this: the amount of storage strength a retrieval builds is inversely related to the current retrieval strength. In plain terms, the harder it is to pull something back (as long as you succeed), the more the act of pulling it back strengthens it for the future. Easy retrievals teach the brain almost nothing because there is nothing to gain. Hard retrievals teach it a lot. Forgetting, within limits, is not the failure of learning. It is the condition that makes the next retrieval worth something.

This is the deep reason desirable difficulties are desirable. They engineer the gap between feeling like you know something and actually having to dig it out. That gap is where storage strength gets built. Remove the gap (cram, reread, get the hint instantly) and you are operating on retrieval strength alone, building a memory with a short shelf life.

What Bjork Got Right

The first thing the desirable-difficulties framework got right is that it inverted a near-universal assumption with hard evidence behind it. The belief that smoother and easier always equals better learning is so intuitive that almost no one questions it. Bjork did not just question it; he produced a body of controlled experiments showing the opposite under specific, definable conditions. That is a rare and valuable thing in behavioral science: a counterintuitive claim that holds up.

The second is the learning-versus-performance distinction itself, which reaches far beyond the classroom. It is a general warning about confusing a short-term signal with a long-term outcome, and it applies anywhere people make decisions based on what they can measure now. Any field that runs quick experiments and optimizes for immediate results is exposed to the exact trap Bjork named. The distinction gave that trap a name and a mechanism.

Third, the framework explained a phenomenon everyone has felt but few could articulate: why the study methods that feel most effective are often the least effective. Rereading and highlighting feel great and do little. Self-testing feels awful and works. By tying the feeling of learning to fluency rather than durability, Bjork explained why our metacognition betrays us so reliably, and why students keep choosing the worse strategy even after being shown the data.

Fourth, and most useful for practitioners, the theory is prescriptive without being vague. It does not just say “challenge is good.” It names specific, testable manipulations (space, interleave, test, generate, vary, delay feedback) and predicts the direction of their effects. That specificity is what lets you build it into a real curriculum or product instead of nodding along to a platitude about the value of hard work.

Where Desirable Difficulties Fall Apart

The theory is strong, but the way it gets repeated in blog posts and conference talks is often sloppy, and the failures are instructive.

The Disfluency Mirage

The most famous overreach is the claim that hard-to-read fonts make material more memorable. A 2011 study by Connor Diemand-Yauman, Daniel Oppenheimer, and Erikka Vaughan reported that students learned material better when it was printed in ugly, disfluent typefaces, and the finding spread everywhere as proof that any difficulty helps. It became a designer’s favorite party fact. The trouble is that follow-up work has produced mixed and frequently null results, including a study pointedly titled “Fortune is fickle,” and no clean direct replication has held up. Visual disfluency is a fragile, contested effect, not a reliable lever. It is the cautionary example of treating “any difficulty” as “desirable difficulty,” which is precisely the mistake the word “desirable” was meant to prevent.

The Expertise Reversal

A difficulty that helps a knowledgeable learner can hurt a novice, and vice versa. Research on what is called the expertise reversal effect shows that instructional supports which help beginners become unnecessary, then counterproductive, as expertise grows. The same applies to difficulties: a beginner overwhelmed by interleaving may need blocking first to form the basic categories before mixing helps. Desirable difficulties are not a universal setting you crank to maximum. They are calibrated to the learner’s current level, and that level keeps moving.

The Undesirable-Difficulty Confusion

The hardest part in practice is telling a desirable difficulty from an undesirable one, because both feel hard. A desirable difficulty is effort that the learner can overcome and that builds durable capability. An undesirable difficulty is effort that exceeds the learner’s resources, or effort spent on something that has nothing to do with the actual learning goal: a confusing interface, broken instructions, irrelevant complexity. Both lower performance. Only one builds learning. Confuse them and you will either strip out the productive struggle or pile on pointless obstacles and call it rigor. The distinction is not in how hard it feels; it is in whether the difficulty is doing the work of learning or just standing in its way.

What’s Really Happening Inside the Brain

The neuroscience under desirable difficulties is the neuroscience of effortful retrieval and consolidation. When you successfully retrieve a memory after some forgetting, you are not playing back a recording. You are reconstructing the memory, and the act of reconstruction reactivates and modifies the underlying neural trace. Each effortful, successful retrieval is thought to reconsolidate the memory in a strengthened, more accessible form, which is why testing yourself does more than rereading: rereading is passive re-exposure, while retrieval is active reconstruction.

Effortful processing also recruits more of the brain. Material that is harder to encode (because it is spaced, mixed, or self-generated) engages deeper and more elaborate processing, forming more retrieval routes back to the memory. A memory with many routes in is far more durable than one with a single, fragile path. Interleaving, in particular, forces the brain to repeatedly notice the differences between categories, building the discriminative representations that let you recognize a new instance you have never seen.

There is also a prediction-error angle that explains why successful effortful retrieval is special. When you strain to recall something and then get it right, your brain registers a gap between the expected difficulty and the successful outcome, and that mismatch is a powerful signal to update and strengthen the trace. An easy retrieval produces no such signal because nothing was at stake. This is the same family of learning signal that drives reward systems: the brain pays attention to outcomes that were not guaranteed. Effortful retrieval manufactures exactly that uncertainty, then resolves it, and the resolution is what gets encoded.

Sleep and time then do the consolidation work that the difficulty set up. Spacing leverages the fact that memory consolidation continues offline, between sessions, especially during sleep, when the hippocampus replays recent experiences and gradually hands stable representations to the cortex. A spaced schedule gives the brain repeated cycles of encoding and consolidation, while a massed schedule crams everything into one window and skips the offline processing that turns fragile traces into stable ones. The difficulty creates the conditions; biology finishes the job in the gaps you left. This is why the gap is not wasted time. The gap is when the learning actually happens.

Desirable Difficulties vs Other Theories

Desirable difficulties sit in a neighborhood of theories about challenge, effort, and learning, and the distinctions sharpen what it actually claims.

Versus Flow. Flow, from Mihaly Csikszentmihalyi, is about the optimal challenge-skill balance that produces total absorption in the moment. It is a theory of the experience of engagement. Desirable difficulties are about the outcome of learning, and they are willing to make the moment feel worse, not better. Flow optimizes the present subjective state; desirable difficulties optimize the future objective skill. A flow state can even be the enemy of learning if its smoothness comes from operating well within your existing ability, where little new gets built.

Versus Cognitive Load Theory. John Sweller’s Cognitive Load Theory gives the precise boundary that desirable difficulties need. Sweller splits mental effort into extraneous load (effort wasted on poor design and irrelevant complexity) and germane load (effort that goes into building mental schemas). An undesirable difficulty is extraneous load. A desirable difficulty is germane load. The two theories are not in tension; they are two halves of the same instruction: remove the effort that does not build understanding, protect the effort that does.

Versus the Testing Effect and Spacing. Retrieval practice and spaced repetition are not rivals to desirable difficulties; they are its two flagship instances. Desirable difficulties is the umbrella principle that explains why both work: they introduce effortful retrieval after forgetting. If you understand storage versus retrieval strength, the testing effect and the spacing effect stop being two separate tricks and become one idea seen from two angles.

Versus Growth Mindset. Carol Dweck’s growth mindset is about how a learner interprets difficulty: as a threat to fixed ability, or as a signal of growth. Desirable difficulties is about which difficulties are worth interpreting that way. The two complement each other neatly. Growth mindset gives a learner the willingness to embrace struggle; desirable difficulties tells you which struggles actually pay off so that willingness is not wasted on pointless ones.

Desirable Difficulties in the Real World

Education and Training

The clearest application is in how we teach. Replace rereading with low-stakes self-testing. Space reviews across weeks instead of cramming before exams. Interleave problem types in homework rather than drilling one kind per chapter. Schools that adopt retrieval practice and spacing routinely see large gains on delayed assessments. The barrier is rarely evidence; it is that these methods make students perform worse and feel worse on the day, so both teachers and learners resist them. The teacher who interleaves gets lower quiz scores this week and better exam scores in June, and only one of those is visible when the curriculum decision gets made.

Product and User Experience Onboarding

This is where the theory collides hardest with prevailing practice. The dominant onboarding pattern is to make the first session as effortless as possible: auto-fill everything, skip the setup, show a five-second tour, get the user to a win immediately. That maximizes activation rate, the in-session performance metric. But if you want the user to actually learn your product, to remember how to do the thing without hand-holding next week, some productive effort in the first session is what builds that durable competence. A guided action the user performs themselves is remembered; a tour they watched is not. The frictionless onboarding that wins the activation test can quietly produce users who never learned the product well enough to stay.

This bites hardest in complex tools, the B2B software a person has to genuinely master to get value. The temptation is to demo every feature in a slick walkthrough so the first session feels impressive and complete. But a walkthrough is a tour, and tours do not transfer. The user who was guided to set up one real workflow with their own data, hitting and recovering from a small snag along the way, will remember how to do it and will do the next one alone. The user who watched a polished overview of twelve features remembers none of them and opens a support ticket the moment they try the first one. The effortful onboarding looks worse in the session-completion metric and produces the only thing that matters in a product people pay to use: a user who can actually operate it.

Marketing and Brand Memory

Advertising lives or dies on memory, and desirable difficulties have a quiet say here too. A small amount of processing effort, a message that makes you complete the meaning yourself rather than spelling everything out, tends to be remembered better than a fully spoon-fed one. A clever line you had to think about for half a second lodges deeper than a literal one. The same logic explains why spaced exposure across time builds stronger brand memory than the same number of impressions crammed into one burst. The media plan is a spacing schedule whether the planner knows it or not.

Healthcare and Behavior Change

Teaching patients to manage a condition or follow a regimen runs into the same wall. A handout read once is forgotten. A patient who has to retrieve and restate the plan, who practices the injection or the breathing exercise rather than watching it, who is followed up across spaced intervals, retains and adheres far better. The effortful version feels less comfortable for both clinician and patient in the appointment, and produces better outcomes after it.

The Elephant in the Room

Here is what almost no growth team will put on a slide. The dashboard measures performance, and the business runs on learning, and those are not the same variable.

Nearly every metric a team watches during a session (completion rate, time-to-value, immediate satisfaction, in-session quiz score, activation) is a performance measure. It is captured during or right after acquisition, which is exactly the window Bjork warns is an unreliable index of durable learning. The outcomes the business actually depends on (a user who retains the skill, remembers the product, forms a habit that survives the novelty fading, can transfer the behavior to a new context) are learning outcomes, and learning is invisible on the acquisition dashboard. It only shows up later, after the measurement window has closed.

So watch what happens when a team runs an A/B test on an onboarding flow. They remove a difficulty: skip the tutorial step, autofill the field, give the hint immediately, cut the lesson shorter and easier. In-session metrics go up, because they just removed effort and effort depresses performance. The test “wins.” They ship it. And they have just stripped out the exact productive struggle that was building durable competence. The metric that improved was performance. The thing they destroyed was learning. And the deficit shows up weeks later as users who churn once the hand-holding stops, or who never got good enough at the product to make it part of their life, by which point no one connects the dots back to the onboarding change that “won.”

This is the uncomfortable conclusion. The frictionless-experience movement, applied without this distinction, is at scale a machine for manufacturing fluent users who do not retain. It is individually rational every time, because every single test is honest about what it measures, and collectively ruinous, because what it measures is the wrong half of the picture. The forces that decide long-horizon behavior are the ones the standard dashboard cannot represent. You can win every test you know how to run and still be training your users to forget you.

How to Apply It with the Octalysis Framework

The Octalysis Framework organizes all of human motivation into eight Core Drives. Desirable difficulties are not a ninth drive. They are a design principle that tells you how much effort to leave in an experience, and that principle plugs directly into the Core Drives that make effort feel worth it.

The Home Drive: Core Drive 2

Desirable difficulty is the raw material of Core Drive 2 (CD2): Development & Accomplishment. Accomplishment is only ever felt against difficulty overcome. The brain prices a reward against the effort it took to earn it, which is why a badge for an effortless task feels hollow and a win that cost you something feels real. This is the deep design consequence the frictionless-UX movement misses: when you strip the difficulty out, you do not just hurt memory, you starve your most reliable White Hat motivation driver. The “win” you hand the user is worthless precisely because you made it free. Bjork’s desirable difficulty and Octalysis CD2 describe the same thing from two sides: the challenge that is hard enough to require real effort and achievable enough to be conquered is both the condition for durable learning and the condition for genuine accomplishment.

The Approach Drives: Core Drives 3, 7, and 1

Core Drive 3 (CD3): Empowerment of Creativity & Feedback is fed by the generation effect. A user who produces the answer, makes the choice, or self-corrects after delayed feedback is in an active, creative loop. A user who is shown everything is passive, and passivity is the death of CD3. Core Drive 7 (CD7): Unpredictability & Curiosity is fed by interleaving and by the forgetting that spacing creates: a mixed sequence is less predictable, and a partly-forgotten item makes the next retrieval feel like a small mystery to solve. Core Drive 1 (CD1): Epic Meaning & Calling is what buys the learner’s permission to endure difficulty at all. People will accept, and even seek out, hard things when they believe the effort matters. Meaning is the budget that pays for productive struggle.

The Danger Drive: Core Drive 8

The line between a desirable and an undesirable difficulty is the exact line where Core Drive 8 (CD8): Loss & Avoidance takes over. Inside the band, difficulty drives CD2 accomplishment. Past it, difficulty becomes a wall, and the dominant feeling flips to frustration, anxiety, and the urge to quit, which is CD8 in its destructive form. Every undesirable difficulty (a confusing interface, an impossible early task, effort spent on irrelevant complexity) is a CD8 trigger wearing the costume of rigor. The design instruction is to scaffold so that difficulty stays in the desirable band where CD2 lives and never tips into the CD8 zone.

The Design Instruction

Here is the move that ties it together, and it is also the test for the undesirable-difficulty confusion. Separate the friction that is in the way from the friction that is the way. Confusing navigation, unclear copy, broken flows, and irrelevant complexity are friction in the way: extraneous load, undesirable difficulty, CD8 fuel. Strip every bit of it. Retrieval, spacing, generation, and productive struggle toward the actual goal are friction that is the way: germane load, desirable difficulty, the raw material of CD2 and CD3. Protect every bit of it. The skill of an S-Tier behavioral designer is not “reduce friction” or “add challenge.” It is telling these two apart, and having the discipline to remove one while defending the other against a dashboard that cannot tell them apart either.

Watch how this plays out in a language-learning app, the most common consumer product built on actual learning. The friction in the way is everything the app does to confuse you: a cluttered home screen, a lesson that does not explain what it wants, a bug in the audio. Strip all of it. The friction that is the way is the part where the app asks you to recall the word before it shows you, makes you type the phrase instead of tapping a multiple-choice answer, spaces your review of old words across days so you have half-forgotten them, and mixes verb tenses in one session instead of drilling one. A product team optimizing for daily-active-users will be tempted to soften every one of those: more multiple choice, less typing, review the words you just saw so the streak feels good, one tense at a time so the lesson feels easy. Each change lifts the in-session numbers. Each change also strips out a desirable difficulty and, with it, a hit of Core Drive 2 accomplishment, leaving a user who feels productive, keeps a streak, and cannot order a coffee in the language after a year. The app that respects the difficulty is harder to love on day one and far harder to quit on day one hundred.

Practical Steps for Designing Productive Difficulty

To put desirable difficulties to work without falling into the traps, run through these in order.

1. Name the learning goal, not the session goal. Write down what you want the person to be able to do next month, not what you want them to complete today. If your goal is a delayed-retention outcome, you have permission to make today harder.
2. Separate the two frictions. List every point of effort in the experience. Tag each one as “in the way” (extraneous, remove) or “the way” (germane, protect). This single audit prevents most of the damage, in both directions.
3. Build in retrieval and spacing. Replace passive review with active recall. Distribute practice across time instead of clustering it. These are the two highest-yield, best-supported difficulties; start here.
4. Calibrate to the learner’s level. Set the difficulty in the band the current learner can overcome, and plan to raise it as they grow. Beginners may need supports that experts must lose. Difficulty is a moving target, not a fixed dial.
5. Measure on a delay, not in the session. Add at least one delayed and one transfer measure to your evaluation. If you only look at in-session performance, you will keep choosing the version that teaches less. The delayed test is your only honest scoreboard.
6. Earn the difficulty with meaning. Make the stakes and the payoff of the effort clear, so the learner accepts the struggle as worthwhile rather than reading it as a broken experience. Hard plus meaningful is engagement; hard plus pointless is churn.

Frequently Asked Questions

What are desirable difficulties in simple terms?

They are ways of practicing that feel harder and slow you down now but make you remember more and apply it better later. Spacing your study, mixing problem types, and testing yourself instead of rereading are the main examples. The difficulty is what triggers durable learning, not an obstacle to it.

What is a desirable difficulty in learning?

A desirable difficulty in learning is a practice condition that makes acquisition feel slower and harder while strengthening long-term retention and transfer. Bjork’s criterion is that the learner must be able to overcome it. Spacing, interleaving, and retrieval practice qualify; confusing instructions and irrelevant obstacles never do.

Which practice methods represent desirable difficulty?

Bjork names five method families that represent desirable difficulty: spacing practice out rather than massing it, interleaving mixed problem types, retrieval practice through self-testing instead of rereading, generation, meaning producing an answer before being shown it, and varying the conditions of practice. An obstacle the learner cannot overcome represents none of them.

Who came up with the idea and when?

Robert Bjork, a memory researcher at UCLA, introduced the term “desirable difficulties” in a 1994 chapter on the training of human beings. He and Elizabeth Bjork have developed the supporting theory, including the New Theory of Disuse, over several decades.

What is the difference between learning and performance?

Performance is what you can observe during or right after training: today’s speed, score, or completion. Learning is the durable change that shows up later on a delayed test or in a new situation. They can move in opposite directions, which is why a method that looks great in the moment can teach less in the long run.

Are desirable difficulties always good?

No. A difficulty is desirable only if the learner can overcome it and it builds real capability. If it exceeds the learner’s resources, or if the effort is spent on irrelevant complexity rather than the actual goal, it becomes an undesirable difficulty that blocks learning. Both feel hard; only one helps.

Why does the disfluent-font claim get criticized?

A 2011 study suggested that hard-to-read fonts improve memory, and it spread widely. Later attempts to replicate it produced mixed and often null results, and no clean direct replication has held up. It is the standard example of mistaking “any difficulty” for a “desirable difficulty,” which is exactly the error the word “desirable” was meant to prevent.

How do desirable difficulties apply to product design?

Most onboarding removes all friction to maximize in-session activation, but some productive effort in the first session is what builds durable competence with a product. A guided action the user performs is remembered; a tour they watched is not. The trick is removing the friction that is in the way while protecting the friction that is the way.

How do desirable difficulties relate to Octalysis?

Desirable difficulty is the raw material of Core Drive 2 (Development & Accomplishment): accomplishment is only felt against difficulty overcome. It also feeds Core Drive 3 (creativity and feedback) and Core Drive 7 (curiosity), while tipping into Core Drive 8 (loss and avoidance) the moment difficulty exceeds what the learner can handle. Frictionless design does not just hurt memory; it starves CD2.

What is the single most useful takeaway?

Stop judging a learning experience by how it feels or performs in the moment. The conditions that feel easiest usually teach the least. Measure on a delay, protect productive struggle, and remove only the friction that has nothing to do with the goal.

Desirable Difficulties Was the Beginning, Not the End

Bjork gave us a precise and uncomfortable truth: for anything you want a person to keep, the experience of acquiring it should not be made as smooth as possible. The struggle, in the right dose, is not the price of the learning. It is the learning. You cannot strip it out and keep the result, any more than you can take the resistance out of weightlifting and keep the muscle.

The reason this matters more now than when Bjork first wrote it is that we have built an entire industry on the opposite premise. Reduce friction, smooth the path, make it effortless: it is the default instruction in every product review and every growth meeting. For a checkout flow, it is correct. For anything the user is supposed to learn, remember, or come to own, it is quietly corrosive, and the dashboard that is supposed to catch the damage is measuring the wrong half of the picture.

The designer who internalizes this gains a rare kind of sight. Where everyone else sees friction to be eliminated, they can tell which friction is in the way and which friction is the way, and they have the discipline to defend the second kind against a metric that wants it gone. That discipline is the difference between a product people complete and a product people keep. Decide what you want your users to still be able to do a month from now, and then have the nerve to make this week a little harder.

References

• Bjork, R. A. (1994). Memory and metamemory considerations in the training of human beings. In J. Metcalfe & A. Shimamura (Eds.), Metacognition: Knowing about Knowing (pp. 185-205). Cambridge, MA: MIT Press.
• Bjork, R. A., & Bjork, E. L. (1992). A new theory of disuse and an old theory of stimulus fluctuation. In A. Healy, S. Kosslyn, & R. Shiffrin (Eds.), From Learning Processes to Cognitive Processes (Vol. 2, pp. 35-67). Hillsdale, NJ: Erlbaum.
• Bjork, E. L., & Bjork, R. A. (2011). Making things hard on yourself, but in a good way: Creating desirable difficulties to enhance learning. In M. A. Gernsbacher et al. (Eds.), Psychology and the Real World (pp. 56-64). New York: Worth Publishers.
• Soderstrom, N. C., & Bjork, R. A. (2015). Learning versus performance: An integrative review. Perspectives on Psychological Science, 10(2), 176-199.
• Schmidt, R. A., & Bjork, R. A. (1992). New conceptualizations of practice: Common principles in three paradigms suggest new concepts for training. Psychological Science, 3(4), 207-217.
• Roediger, H. L., & Karpicke, J. D. (2006). Test-enhanced learning: Taking memory tests improves long-term retention. Psychological Science, 17(3), 249-255.
• Kornell, N., & Bjork, R. A. (2008). Learning concepts and categories: Is spacing the “enemy of induction”? Psychological Science, 19(6), 585-592.
• Cepeda, N. J., Pashler, H., Vul, E., Wixted, J. T., & Rohrer, D. (2006). Distributed practice in verbal recall tasks: A review and quantitative synthesis. Psychological Bulletin, 132(3), 354-380.
• Rohrer, D., & Taylor, K. (2007). The shuffling of mathematics problems improves learning. Instructional Science, 35(6), 481-498.
• Diemand-Yauman, C., Oppenheimer, D. M., & Vaughan, E. B. (2011). Fortune favors the bold (and the italicized): Effects of disfluency on educational outcomes. Cognition, 118(1), 111-115.
• Kalyuga, S. (2007). Expertise reversal effect and its implications for learner-tailored instruction. Educational Psychology Review, 19(4), 509-539.
• McDaniel, M. A., & Butler, A. C. (2011). A contextual framework for understanding when difficulties are desirable. In A. S. Benjamin (Ed.), Successful Remembering and Successful Forgetting (pp. 175-198). New York: Psychology Press.
• Chou, Y. (2015). Actionable Gamification: Beyond Points, Badges, and Leaderboards. Octalysis Media.

Related Reading

• Retrieval Practice and the Testing Effect: Why Recall Beats Review
• Spaced Repetition and the Forgetting Curve: Learning That Lasts
• Cognitive Load Theory: The Difference Between Effort That Builds and Effort That Blocks
• Growth Mindset: How a Learner Reads the Meaning of Difficulty
• The Octalysis Framework: The 8 Core Drives of Human Motivation
• The Behavioral Framework Library: Every Model in One Place

WOULD YOU LIKE YU-KAI CHOU TO WORK WITH YOUR ORGANIZATION?