Desirable Difficulties: Why Effortless Learning Deceives and Memory Lives on Effort

The Hook: The Best Learning Feeling Is Often the Worst Teacher

Picture two students preparing for the same exam. The first reads her notes through four times in a row, highlights the key passages in yellow, and reads them once more. By the end of the evening the material feels familiar, smooth, mastered — a good, reassuring feeling. The second reads the notes only once, closes them, and tries to write down from memory what she has retained. It stutters; she gets stuck on gaps; she is annoyed by what she can no longer recall. By the end of the evening she feels less confident than the first.

A week later, both sit the exam. And now everything reverses: the second student, the one who felt worse while studying, performs markedly better. This is neither an anecdote nor a one-off but one of the most robust and most frequently replicated findings in cognitive psychology. It has a name that sounds paradoxical at first: desirable difficulties, a term coined by the memory researcher Robert Bjork.

The core message is unsettling and liberating at once: the feeling of learning well and actual learning are not merely different — they often run directly against each other. Methods that feel fluent and effortless (rereading, highlighting, summarizing) produce an illusion of competence yet leave surprisingly little durable knowledge behind. Methods that feel laborious, halting, and sometimes frustrating (testing yourself, inserting gaps, mixing topics) build a memory that survives weeks and months.

Why should this matter to you — beyond school? Because you learn your whole life: a new programming language, a certification topic, the architecture of an unfamiliar system, a foreign language, the material for the next exam. The time you invest is scarce. And the cognitive psychology of the last hundred and forty years has worked out, rather precisely, which methods convert that time into durable knowledge — and which merely convert it into a good but deceptive feeling. This article takes you the full distance: from Ebbinghaus's forgetting curve through the testing effect and the spacing effect to Bjork's theory of storage and retrieval strength — and ends with a concrete, practically usable procedure.

Part 1: The Forgetting Curve — Memory as a Leaky Vessel

Hermann Ebbinghaus, His Own Lab Rat

Before you can understand how to retain durably, you have to understand how you forget. And the history of scientific forgetting begins with a German philosopher who, in the 1880s, turned himself into his own lab rat. Hermann Ebbinghaus set out to prove that higher mental processes — until then deemed unmeasurable — could be studied experimentally and quantitatively. His tools were roughly 2,300 nonsense syllables (consonant–vowel–consonant, such as "WID," "ZOF"), purpose-built to strip out the influence of pre-existing meaning and association. He learned lists of them until he could recite them flawlessly, then measured, after varying intervals, how much repetition was needed to master them again. This "method of savings" is ingenious: what is saved in practice is precisely what remains of the original learning.

The result was the famous forgetting curve: forgetting is not linear but steep at first and then ever flatter. Most of what is freshly learned is lost in the first hours and days; what survives this first wave then remains surprisingly stable. Memory, figuratively, is a leaky vessel that leaks quickly at first and ever more slowly thereafter.

A 130-Year-Old Curve, Successfully Replicated

Here is an important point for anyone who — like Sven — values scientifically secured statements: Ebbinghaus's finding is not a historical curiosity that would fail a replication test. On the contrary. In 2015, Jaap Murre and Joeri Dros published a careful replication of Ebbinghaus's original experiment in PLOS ONE. A single subject spent about 70 hours learning lists and relearning them after 20 minutes, 1 hour, 9 hours, 1 day, 2 days, or 31 days. The result was a successful confirmation of the classic 1880 curve — with one interesting subtlety: the curve is not perfectly smooth but probably shows a slight upward jump around the 24-hour point, which today is associated with the memory-consolidating effect of sleep.

The lesson of the forgetting curve is not resignation but strategy. If forgetting happens mostly early, then the timing of repetition is decisive. A refresher exactly when the curve begins to drop "repairs" memory most effectively — and flattens the next forgetting curve. This is precisely where the two most powerful learning principles we know come in: retrieving and spacing.

Part 2: The Testing Effect — Retrieving Is Not Measuring, It Is Learning

The Central Reversal

For centuries a test was regarded as a pure measuring instrument: you learn, and then an exam gauges how much stuck. The test itself, ran the tacit assumption, adds nothing to the knowledge; it merely reads it off, like a thermometer reading temperature. This assumption is wrong — and its refutation is one of the most consequential insights of modern learning research.

The testing effect (also the retrieval practice effect) states: the act of actively retrieving something from memory changes memory itself — it strengthens the trace durably, often more than re-studying the same material does. Retrieval is not a read-out but a learning event in its own right. Put differently: every time you successfully remember something, you make the next act of remembering more likely.

The Key Experiment: Roediger and Karpicke 2006

The modern renaissance of this effect is closely tied to an experiment by Henry Roediger and Jeffrey Karpicke (2006), published in Psychological Science. The design is instructive because it pits against each other exactly the two camps that opened this article.

Students read short factual passages (about 250 words). They were then split into groups. One group studied the text repeatedly. The other group tested itself after reading by writing down, without the text in front of them, what they had retained — free recall, with no feedback at all. The final retention test was then administered after varying intervals: after 5 minutes, after 2 days, or after 1 week.

The result is the famous crossover:

• On the test after 5 minutes, the studying group performed better than the testing group. Rereading works in the short term — which is exactly what feeds the illusion.
• On the test after one week, the relationship reversed dramatically: those who had tested themselves while learning retained markedly more than those who had merely reread.

In Roediger and Karpicke's second experiment, which compared different numbers of study and test trials, the effect can be made vivid in numbers. Pure repeated study (four study trials) produced the highest recall after 5 minutes (around 83%) but collapsed to about 40% after a week. The repeated-retrieval condition (one study trial, then three test trials) was lower after 5 minutes (around 71%) — yet retained far more after a week, at about 61%. The method that "felt" weaker in the short term was by far the better one in the long term. These numbers come from that single experiment and should be read as an illustration of the pattern, not as universal constants; the pattern itself — the crossover — has been replicated many times.

Retrieval Beats Even Elaborate Studying

A common objection runs: "Testing yourself may be fine for rote memorization, but real understanding requires more sophisticated methods." That, too, was tested. Karpicke and Blunt (2011) pitted retrieval practice directly against concept mapping in Science — an elaborate, much-praised method in which learners process material into networked concept diagrams. The result: when learning from factual texts, retrieval practice produced more meaningful learning on a test one week later than concept mapping did — and not only for plain factual knowledge but also for comprehension questions that required inferences. Remarkably, the students themselves expected concept mapping to perform better. They were wrong about their own method.

Why Does Retrieval Work?

The cognitive explanation is plausible and well supported: successful retrieval is a demanding reconstruction process. The brain has to find the right cues, reactivate the trace, suppress competing memories, and reassemble the result. This work leaves traces: it broadens the number of retrieval routes, consolidates the connections, and anchors the knowledge more deeply in the existing web. Rereading, by contrast, is passive — the brain comfortably "recognizes" the material without traveling the strenuous reconstruction path. Recognition is easy and feels good; reconstruction is hard and teaches. (Even unsuccessful retrieval followed by feedback can, incidentally, be useful — the attempt alone primes memory for the correction.)

Part 3: The Spacing Effect — Why Gaps Build Memory

The Second Great Principle

If the testing effect concerns the how of learning (retrieve rather than reread), then the spacing effect (the distributed-practice effect) concerns the when. Its claim is as simple as it is counterintuitive: the same amount of study time produces markedly more durable knowledge when it is spread out over time rather than bundled into a single session. Two hours of learning in one block (massed practice, "cramming") are far inferior, for memory, to four times thirty minutes on different days — even though the time spent is identical.

This too is not a conjecture but is hard-backed by meta-analysis. Cepeda, Pashler, Vul, Wixted, and Rohrer (2006) assembled, in a large quantitative synthesis, 839 comparisons from 317 experiments. The finding was unambiguous and massive: distributed practice beats massed practice across an enormous range of materials and conditions.

The Decisive Twist: There Is No Single "Optimal" Interval

The truly deep insight of the meta-analysis is subtler and decisive for practice. There is no single ideal gap between two learning sessions. Rather, the optimal gap depends on how long the knowledge needs to last. The inter-study interval (the gap between repetitions) and the retention interval (the time until the test) act together: the further in the future the test lies, the larger the gaps between repetitions should be.

A practical rule of thumb derived from this research: the optimal gap between two study sessions is roughly 10 to 30 percent of the period over which you want to retain the knowledge. If you want to retain something for a week, a gap of about a day is favorable; if you want to retain it for a year, gaps of weeks make sense. The important thing is the direction of the insight: anyone studying for a distant exam and stacking the repetitions too tightly is leaving potential on the table.

Why Does Spacing Work?

There are several mutually complementary explanations. One traces back to the theory of stimulus fluctuation: at each study session the internal and external context (mood, surroundings, stream of thought) is slightly different; distributed practice therefore links the material to more varied cues, which later opens up more retrieval routes. A second explanation ties directly to the forgetting curve: a repetition has its strongest effect when the knowledge has already faded a little — then reactivation is more effortful and therefore more instructive. Immediate repetition meets a still-fresh memory and does almost no work. Here the circle closes back to Bjork's theory, which brings both principles — retrieving and spacing — under a single roof.

Part 4: Storage Strength and Retrieval Strength — Bjork's Elegant Theory

Two Memory Properties Instead of One

With their "New Theory of Disuse," Robert and Elizabeth Bjork supplied a conceptual tool that orders the seemingly paradoxical findings with surprising clarity. Their core idea: every memory possesses not one but two independent strengths.

Storage strength measures how deeply and how well-networked a content is anchored in the overall fabric of knowledge. The decisive property: storage strength can (according to this theory) only increase — once deeply anchored, something stays anchored. It does not decline.

Retrieval strength measures how easily a content is accessible right now. It is fleeting: it shoots up after learning but falls quickly again without use, and depends heavily on the cues currently available.

This is precisely where all the confusion has its root. When you have just read something four times, your retrieval strength is high — the material feels present and mastered. But that feeling says almost nothing about storage strength, that is, about whether the knowledge will still be there in a week. We chronically confuse the easily felt retrieval strength with the genuinely important storage strength. That is the psychological mechanism behind the illusion of competence.

The Punchline: Difficult Retrieval Builds More Storage Strength

From the theory follows an elegant, almost poetic prediction: the gain in storage strength is greater the lower the retrieval strength was at the moment of successful retrieval. Put differently: remembering something laboriously and against the resistance of forgetting consolidates it more strongly than when it comes to mind effortlessly. Easy retrieval teaches little; hard (but successful) retrieval teaches a lot.

This explains both great principles at a stroke. The testing effect works because a test actively draws on retrieval strength rather than merely raising it passively. The spacing effect works because gaps deliberately let retrieval strength fall — the next repetition then meets a lower retrieval strength and accordingly produces more storage strength. Both methods make learning harder in the moment and feel worse — and that is exactly why they work. That is the definition of a desirable difficulty: an effort that promotes long-term retention rather than hindering it.

Forgetting Is Not a Defect but a Function

A beautiful consequence of this theory: forgetting no longer appears as a mere deficiency but as a meaningful mechanism. By letting the retrieval strength of unused contents decline, the system keeps what is currently relevant accessible and pushes the outdated into the background. And — counterintuitively — a bit of forgetting is what creates the very precondition for the next repetition to have a powerful effect. Someone who never forgets cannot benefit from the labor of remembering anew.

Part 5: The Illusion of Competence — Why We Love the Wrong Methods

We Are Poor Judges of Our Own Learning

If retrieving and spacing are so superior — why does almost no one do them spontaneously? The answer lies in a stubborn metacognitive deception. In surveys, students overwhelmingly report rereading as their main strategy; only a minority test themselves. The reason is not laziness but a genuine perceptual error: the fluency of recognition during rereading feels like understanding. Roediger and Karpicke called this the illusion of competence — one mistakes the momentarily high retrieval strength for durable knowledge.

Especially insidious: the students in the experiments consistently predicted that rereading would benefit them more in the long term than self-testing would — the exact opposite of what then occurred. Our internal model of how learning works is systematically biased. We trust the feeling, and the feeling rewards the wrong thing.

Here the topic touches a deeply epistemological question that stands at the center of Three Pages Against 2,000 Years: The Gettier Problem and What Knowledge Really Is: when do you really know something, and when do you merely have the feeling of knowing it? The illusion of competence is, in a sense, the psychological variant of the Gettier problem: a state that feels from the inside like secure knowledge but does not withstand outside examination. The self-test is the tool that exposes this gap — it confronts the feeling of knowing with the fact of being able.

The Role of Errors — and Why an Error-Friendly Culture Helps

To allow a desirable difficulty is to allow errors. Whoever tests themselves will produce gaps and mistakes — that is not a side effect but the mechanism itself. Studies suggest that even generating wrong answers followed by correction improves later retention. That, however, presupposes an inner stance that experiences one's own error not as a threat but as information. At the level of teams and organizations, this is precisely the bridge to Psychological Safety: The Hidden Force Behind High-Performing Teams: in a climate that treats mistakes as a learning opportunity, people are more willing to expose themselves to the uncomfortable test, to reveal knowledge gaps, and thereby actually to learn. Where mistakes are punished, learners flee into the comfortable but ineffective method of mere rereading.

Part 6: From Principle to Practice — A Toolkit for Durable Learning

The research translates into a manageable procedure. The following table ranks the most important techniques by their empirically demonstrated effectiveness — the assessment follows in essence the large review by Dunlosky, Rawson, Marsh, Nathan, and Willingham (2013), which systematically compared ten common learning techniques.

From this follows a concrete procedure that can be applied to almost any material — including what you learn professionally:

Read or work through a section once attentively. Then close it and retrieve from memory what stuck — write it down, say it aloud, or explain it to a (real or imaginary) person. Only afterward check against the source and correct. Distribute the repetitions over several days rather than cramming everything into one session, and deliberately let some time pass so that retrieval stutters a little. Mix related topics instead of working through each in isolation in a block. And above all: distrust the feeling of ease. If learning feels effortless, you are probably learning little just then. The stuttering is not the failure of learning — it is the learning.

A note on the honesty of the evidence: the testing effect is strongest and most broadly documented for clearly testable knowledge (facts, concepts, relationships). For very complex, ill-structured, or highly transfer-dependent contents, the evidence is more mixed, and good retrieval exercises are harder to design there. I am of the opinion that this is not an argument against the principle but an invitation to make the testing demanding — with questions that require application and inference, not mere reproduction. How artificial neural networks store their "contents" in distributed representations, and how hard it is to tell from the outside what is really "known," is incidentally a fascinating technical mirror of the same problem — to be read in The Ghost in the Machine: How to Read a Neural Network From the Inside.

The Central Takeaway

The central, transferable insight behind all these findings is a single, uncomfortable, but extraordinarily useful truth:

Learning success and the feeling of learning are two different things — and the most effective methods feel the worst in the moment. Rereading, highlighting, and effortless recognition produce a high but fleeting retrieval strength and with it the warm feeling of mastery; yet they leave little storage strength behind. Testing yourself and distributing learning over time feels more laborious and more uncertain — and it is precisely that effort which is the price minted into durable knowledge.

Apply this to your own practice. The next time you really want to retain something — a new technology, a certification topic, the structure of an unfamiliar system — do not ask yourself "Have I understood it?" but submit to the harder test: "Could I, right now, with the book closed, reconstruct it from memory, and again in a week?" Deliberately build retrieval and spacing into your learning process, even when your feeling protests. The feeling is a poor advisor here. "Deliberately make it a bit harder for yourself" is more than a study trick — it is a stance toward your own knowledge that makes the difference between fleeting familiarity and resilient capability.

Reflection Question

Think of something you have learned in recent weeks and consider professionally "mastered" — a library, a concept, an architecture. How did you learn it: by rereading and looking things up, or by actively reconstructing from memory? And now the honest test: if you had to write down the core idea right now, without looking, or explain it to a colleague — how much would actually come out, and how much is just the pleasant feeling of familiarity? Which one thing that you are currently learning would you approach differently if you made it a principle to distrust the feeling of ease?

Cross-References in the Vault

• Three Pages Against 2,000 Years: The Gettier Problem and What Knowledge Really Is – The "illusion of competence" is the psychological sister of the Gettier problem: a state that feels from the inside like knowledge but does not withstand outside examination. The self-test is the tool that pulls the two apart.
• Psychological Safety: The Hidden Force Behind High-Performing Teams – Desirable difficulties demand that errors be allowed. An error-friendly, psychologically safe culture is the social precondition for people to expose themselves at all to the uncomfortable but instructive test.
• The Ghost in the Machine: How to Read a Neural Network From the Inside – How artificial networks store their "knowledge" in distributed representations, and how hard it is to tell from the outside what is really mastered, is a technical mirror of the question of storage versus retrieval strength in human memory.

Sources

• Hermann Ebbinghaus (1885): Über das Gedächtnis. Untersuchungen zur experimentellen Psychologie (On Memory: A Contribution to Experimental Psychology). Duncker & Humblot, Leipzig. (Founding of the forgetting curve and the method of savings.)
• Jaap M. J. Murre & Joeri Dros (2015): Replication and Analysis of Ebbinghaus' Forgetting Curve. PLOS ONE 10(7): e0120644. https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0120644
• Henry L. Roediger III & Jeffrey D. Karpicke (2006): Test-Enhanced Learning: Taking Memory Tests Improves Long-Term Retention. Psychological Science 17(3), pp. 249–255. https://journals.sagepub.com/doi/10.1111/j.1467-9280.2006.01693.x
• Jeffrey D. Karpicke & Janell R. Blunt (2011): Retrieval Practice Produces More Learning than Elaborative Studying with Concept Mapping. Science 331(6018), pp. 772–775. https://pubmed.ncbi.nlm.nih.gov/21252317/
• Nicholas J. Cepeda, Harold Pashler, Edward Vul, John T. Wixted & Doug Rohrer (2006): Distributed Practice in Verbal Recall Tasks: A Review and Quantitative Synthesis. Psychological Bulletin 132(3), pp. 354–380. https://pubmed.ncbi.nlm.nih.gov/16719566/
• John Dunlosky, Katherine A. Rawson, Elizabeth J. Marsh, Mitchell J. Nathan & Daniel T. Willingham (2013): Improving Students' Learning With Effective Learning Techniques: Promising Directions From Cognitive and Educational Psychology. Psychological Science in the Public Interest 14(1), pp. 4–58. https://journals.sagepub.com/doi/abs/10.1177/1529100612453266
• Elizabeth L. Bjork & Robert A. Bjork (2011): Making Things Hard on Yourself, But in a Good Way: Creating Desirable Difficulties to Enhance Learning. In: Psychology and the Real World. https://bjorklab.psych.ucla.edu/wp-content/uploads/sites/13/2016/04/EBjork_RBjork_2011.pdf

← All articles