Desirable Difficulties: Why Harder Study Methods Produce Better Learning

In 1994, Robert Bjork coined a phrase that has since become one of the most important concepts in learning science: desirable difficulties. These are conditions that slow apparent learning in the short term — making performance worse during study — while producing superior long-term retention and transfer.

The implication is uncomfortable. Most of what students do to make studying feel productive (re-reading, blocked practice, smooth fluent review sessions) is exactly wrong. Most of what produces genuine learning feels frustrating and slow.

Understanding why is not an academic exercise. It directly changes which study methods you should use.

The core paradox: performance vs. learning

Bjork's key insight was that current performance during practice is a poor indicator of learning. Learning is defined as durable improvement in long-term retention and ability to apply knowledge in new contexts. Performance is what happens now, in this session, on today's quiz.

These can be — and often are — dissociated. Conditions that maximise within-session performance can minimise long-term learning. Conditions that feel unproductive during a session can produce substantially better long-term outcomes.

The most documented example: re-reading a text chapter immediately after reading it produces high performance on an immediate comprehension test. It produces nearly zero improvement in retention at one week compared to a single read. A single retrieval practice session after the first read produces 50% better retention at one week (Karpicke & Roediger, 2008). The re-reading felt productive. The retrieval practice felt harder. The learning was radically different.

The four main desirable difficulties

1. Spacing (distributed practice)

What it is: Reviewing material at intervals that allow partial forgetting before each review, rather than massing all study in a single session.

Why it is desirable: When retrieval happens after a delay, the memory trace has partially faded and must be actively reconstructed. This reconstruction effort is what strengthens the trace. Immediate re-reading bypasses this reconstruction entirely — the trace is too fresh to require effort.

Evidence: Cepeda et al.'s (2006) meta-analysis of 254 studies found spaced practice produces 1.5–3× better retention than massed practice at delays of one week or more.

The difficulty: Spaced study feels less productive because within any given session, your recall is worse than if you had studied the material recently. The poor within-session recall is the mechanism, not a problem to be solved.

For the full treatment, see What Is Spaced Repetition?

2. Interleaving

What it is: Mixing different topics or problem types within a study session rather than completing all of one type before moving to the next.

Why it is desirable: Interleaving forces discriminative judgement on every problem — you must identify which type of problem this is and which method applies before executing. This identification step is precisely what exams require. Blocked practice trains execution but not identification.

Evidence: Rohrer and Taylor (2007) found interleaved maths practice produced 43% better performance at a 4-week delayed test despite no difference at immediate test. Kornell and Bjork (2008) found interleaved study produced 78% better performance on a one-week delayed recognition test.

The difficulty: During interleaved study, performance on each problem feels lower because the previous problem used a different method. Students consistently perceive interleaving as less productive than blocking — while performing better on delayed tests.

For the full treatment, see Interleaving Study Technique.

3. Testing effect (retrieval practice)

What it is: Attempting to retrieve information from memory — rather than re-reading, reviewing, or re-exposing to information — as the primary form of review.

Why it is desirable: Retrieval is not simply a measurement of learning — it is itself a learning event. The act of searching for and reconstructing a memory strengthens that memory's future accessibility. Each successful retrieval makes the next retrieval easier. Reading the same notes for the fourth time strengthens a trace that is already familiar; retrieving it strengthens a trace that must be worked for.

Evidence: Roediger and Karpicke (2006) found that students who took three tests after reading a passage retained 61% of material one week later; students who restudied the passage three times retained 40%. The testing group scored 50% higher despite spending less time with the material.

The difficulty: Retrieval practice feels frustrating, particularly when retrieval fails (which is common and useful). Re-reading produces a comfortable feeling of familiarity. This familiarity is a poor proxy for retrievability.

The key desirable difficulty tools on this site: Spaced Repetition Flashcards and Active Recall course.

4. Generation effect

What it is: Producing information yourself — generating an answer, creating an example, reconstructing a concept — rather than receiving it ready-made.

Why it is desirable: Slamecka and Graf (1978) demonstrated that participants who generated words from partial cues (e.g., completing "syn___" to produce "synonym") remembered those words significantly better than participants who read the complete words. Generation requires deeper processing because the item must be retrieved or constructed from existing knowledge rather than simply encoded from the input.

Evidence: The generation effect is robust across vocabulary, mathematical problem-solving, factual recall, and conceptual understanding. It applies whether the generation succeeds or fails — attempting to produce an answer before consulting the source improves retention even when the attempt is incorrect.

Applications:

• Attempt to recall the main points of a chapter before reading it (pre-testing)
• Generate your own examples for abstract concepts before reading provided examples
• Reconstruct a topic outline from memory before comparing to your notes
• Answer practice questions before reviewing how they should be answered

The difficulty: Generation attempts are slower than reading. They produce more errors. They feel less productive. They consistently produce better retention.

Distinguishing desirable from undesirable difficulties

Not all difficulty in studying is desirable. Bjork's framework distinguishes between:

Desirable difficulty: Slows performance but strengthens learning (spacing, interleaving, retrieval, generation)

Undesirable difficulty: Slows performance without strengthening learning — or prevents engagement with the material entirely (unclear instructions, insufficient prior knowledge, distracting environments, studying material too advanced for the current knowledge base)

The diagnostic question: "Is this difficulty making my brain work harder on the right thing?" If yes, it is likely desirable. If the difficulty prevents engagement with the material at all (you genuinely cannot understand what you are reading), it is undesirable and should be addressed by choosing more appropriate material or seeking assistance.

Applying desirable difficulties in a study session

A session structured around desirable difficulties looks different from a typical revision session:

For the interleaving course with practical exercises, see Interleaving. For the spaced repetition system, see Spaced Repetition.

References

• Bjork, R.A. (1994). Memory and metamemory considerations in the training of human beings. In Metacognition, 185–205. MIT Press.
• Cepeda, N.J., et al. (2006). Distributed practice in verbal recall tasks. Psychological Bulletin, 132(3), 354–380.
• Karpicke, J.D., & Roediger, H.L. (2008). The critical importance of retrieval for learning. Science, 319(5865), 966–968.
• Kornell, N., & Bjork, R.A. (2008). Learning concepts and categories. Psychological Science, 19(6), 585–592.
• Roediger, H.L., & Karpicke, J.D. (2006). Test-enhanced learning. Psychological Science, 17(3), 249–255.
• Rohrer, D., & Taylor, K. (2007). The shuffling of mathematics problems improves learning. Instructional Science, 35, 481–498.
• Slamecka, N.J., & Graf, P. (1978). The generation effect. Journal of Experimental Psychology: Human Learning and Memory, 4(6), 592–604.