Knowledge Concepts
Desirable difficulties are learning conditions that slow apparent progress in the short term but produce stronger long-term retention — like spacing
Desirable difficulties are learning conditions that make practice harder and feel less productive in the moment — but produce significantly better long-term retention and transfer. Spacing out practice, mixing topics, and testing yourself all feel harder than blocked practice and re-reading. They also produce learning that actually sticks. The difficulty is the mechanism, not a side effect.
The concept inverts a common intuition: easier practice doesn't mean better learning. The conditions that feel most productive are often the least effective.
Robert Bjork, a cognitive psychologist at UCLA, coined the term "desirable difficulties" in a 1994 paper ("Memory and Metamemory Considerations in the Training of Human Beings"). Bjork and his collaborators observed that many training practices that produce strong short-term performance actually produce weak long-term retention — and vice versa.
The paradox arises because learners use fluency as a proxy for learning: if something feels easy and comes quickly, it must be learned. But fluency during practice is often produced by conditions (massed practice, re-reading, blocked study) that don't transfer to durable memory. Conditions that feel harder (spaced practice, interleaving, self-testing) produce slower apparent progress but far better retention when tested a week or a month later.
Subsequent research by Bjork, Elizabeth Bjork, and researchers including John Dunlosky (whose 2013 meta-analysis of learning techniques is foundational) confirmed and extended these findings.
Spaced practice (spacing effect): Distributing practice across time ("study for 30 minutes on three separate days") produces better retention than massing the same total time in one session ("study for 90 minutes in one sitting"). The forgetting that happens between sessions is not a problem to eliminate — it's what makes the next study session effective.
Retrieval practice (testing effect): Testing yourself on material produces better retention than re-reading it, even if the test is imperfect and the answers are sometimes wrong. The generation effect is the mechanism: producing information from memory strengthens the memory trace more than recognizing it when presented.
Interleaving: Mixing different problem types or topics within a study session ("practice calculus problems, then history terms, then programming concepts, then more calculus") produces better long-term performance than blocked study ("all calculus, then all history"). Interleaving is harder and more effortful — which is why it works.
Generation (making information before receiving it): Attempting to answer a question, define a term, or derive a result before seeing the answer produces better retention than reading the answer directly — even when the attempt is incorrect.
Varying conditions of practice: Practicing in different environments, with different problem variations, and at different times produces more flexible, transferable learning than always practicing under identical conditions.
The key mechanism: difficulty in practice forces deeper processing. When you space practice, you have to reconstruct partially forgotten material — that reconstruction is harder than re-reading, and the reconstructive effort strengthens memory. When you interleave topics, you have to identify which strategy applies to which problem — harder than blocked practice where the strategy is always the same, and the identification effort builds flexible knowledge.
The cognitive science term for the underlying process is "elaborative encoding" — the formation of rich, associative memory traces that have many connections to existing knowledge and can therefore be retrieved from many different cues. Easy conditions (re-reading, massed practice, blocked study) produce shallow encoding. Difficult conditions (retrieval, spacing, interleaving) produce elaborative encoding.
The feeling of learning is not the same as learning. What feels smooth, fluent, and productive may be producing weak, fast-decaying memory. What feels slow, effortful, and uncertain may be producing exactly the durable encoding that matters.
A medical student is studying pharmacology.
Undesirable easy approach: Study all cardiovascular drugs for three days. Read notes repeatedly. Complete a cardiovascular practice test. Move on to respiratory drugs. Study those. Complete a respiratory practice test.
Apparent result: high scores on each chapter test. Feels productive. Actual result one month later: poor performance on integrated exam that mixes cardiovascular and respiratory drug questions. Can't distinguish which mechanism applies to which drug class under exam conditions.
Desirable difficult approach: Study both cardiovascular and respiratory drugs, interleaved. Each day: mix question types across both categories. Space the review sessions: day 1, day 3, day 7, day 14, day 30. Use flashcards (active recall) rather than re-reading.
Apparent result: lower scores during practice sessions. Feels slower and more frustrating. Actual result one month later: strong performance on integrated exam. Can apply knowledge flexibly to new drug presentations.
Not all difficulties improve learning. Bjork distinguishes:
| Difficulty type | Example | Effect on retention |
|---|---|---|
| Desirable | Spacing practice, interleaving, retrieval | Improves long-term retention |
| Undesirable | Unclear instructions, confusing formatting, irrelevant complexity | Reduces learning (extraneous cognitive load) |
| Context-dependent | Reducing font legibility | Mixed — some inconvenience may force processing; too much impairs it |
Desirable difficulties share key properties: they impede performance during practice but improve performance during subsequent retrieval. The student who interleaves topics performs worse during interleaved study than during blocked study — but performs better on a later test. The difficulty impedes performance without impairing learning; it produces learning by demanding more effortful processing.
Undesirable difficulties impede both performance and learning. Poorly organized notes, unclear explanations, and irrelevant complexity don't produce the processing that strengthens memory — they just waste cognitive capacity.
Step 1 — Replace re-reading with retrieval. When reviewing material: close your notes and try to recall the key points before checking. This adds the desirable difficulty of retrieval practice to your review sessions.
Step 2 — Space your practice. Instead of studying a topic intensively on one day, distribute the same total time across multiple sessions separated by days or weeks. Your calendar becomes a learning tool: schedule topics for review before you'd forget them.
Step 3 — Interleave topics. When working through problems or reviewing content: mix topics within each session rather than working through one topic exhaustively before moving to the next. This is more uncomfortable but produces more flexible retention.
Step 4 — Attempt before receiving. Before reading an explanation: try to work out the answer yourself. Before watching a tutorial: try to figure out the approach. The attempt, even if wrong, primes deeper processing of the correct information.
Step 5 — Accept the feeling of not knowing. The feeling of effortful, uncertain learning — not knowing whether you've got it yet — is a signal that desirable difficulties are working. The feeling of smooth, fluent review is often a signal that you're not learning anything new.
The capturing paradox: Saving information is easy (one click in WebSnips). Processing information — writing your own summary, testing yourself on the content, returning to it days later — is harder. The desirable difficulties framework says: the hard part is the learning. The capture is just the starting condition.
Designing desirable difficulties into your reading workflow:
This workflow adds desirable difficulties to the natural reading process. It's slower than reading and immediately moving on — and produces dramatically better retention.
"Harder always means better." No. Only difficulties that force deeper processing — retrieval, spacing, interleaving — are desirable. Irrelevant complexity, poor organization, and confusing instructions are undesirable difficulties that impede learning without improving encoding.
"If I feel like I'm learning, I'm learning." The opposite is often true. Fluency (reading smoothly, recognizing information easily) is a poor indicator of durable encoding. Difficulty (struggling to recall, not knowing yet) is often the signal of effective processing.
"Interleaving only works for problems with a clear solution." Interleaving applies to conceptual knowledge as well as procedural skills. Mixing topics in note-review sessions — cycling through different subjects in the same session — produces the same benefit as mixing problem types.
Spaced repetition: The systematic scheduling of practice at increasing intervals — the operational implementation of the spacing desirable difficulty.
Active recall: Retrieval practice — the testing desirable difficulty implemented as a habit.
Generation effect: The specific mechanism by which attempting to produce information (even incorrectly) produces stronger encoding than receiving it.
Cognitive load: Desirable difficulties increase germane cognitive load (productive processing) without increasing extraneous load (navigating confusing presentation).
How do I know if a difficulty is desirable or just frustrating? Desirable difficulties make practice harder but leave you with better retention when tested later. Undesirable difficulties impede performance without improving later test performance. The practical test: do you perform better on a delayed test after the difficult practice than after easier practice? If yes, the difficulty was desirable.
Are desirable difficulties more important for some kinds of learning than others? The spacing and retrieval effects appear across virtually all learning types and age groups studied. Interleaving has the strongest evidence for procedural skills (math problem-solving, medical diagnosis, classification tasks) but also applies to conceptual learning. Generation applies most strongly to initial encoding of facts and concepts.
I'm preparing for an exam in two days. Should I still use spaced practice? For a short-term exam, the optimal approach depends on your timeline. Spacing requires time for gaps. If you only have two days: retrieval practice (self-testing) produces the best short-term gains. If you have a week or more: space your review sessions and use retrieval practice within each session.
Desirable difficulties reframe the experience of learning: the struggle, the uncertainty, the slowness are not indicators that something is wrong — they're indicators that something effective is happening. The conditions that feel most productive (smooth re-reading, blocked practice, massed study) produce the weakest retention. The conditions that feel least productive (retrieval, spacing, interleaving) produce the strongest. Understanding this doesn't make the difficulty disappear — but it makes you willing to seek it out rather than avoid it.
See also: The Ultimate Guide to Web Clipping.
More WebSnips articles that pair well with this topic.
Knowledge transfer is the deliberate process of moving knowledge from where it exists — an individual, team, or system — to where it is needed, in a form
A browser extension is a small software add-on installed in a web browser that adds features or modifies behavior — blocking ads, saving passwords
A content calendar is a planning tool that schedules what content will be published, when, where, and by whom — turning a content strategy from vague
A context window is the maximum amount of text an AI language model can process in a single interaction — everything in the prompt, the conversation
A knowledge silo is a condition where knowledge, information, or expertise is isolated within a team, department, or individual — inaccessible to others
A large language model (LLM) is a neural network trained on massive amounts of text to predict and generate language.