Efficient Human Learning

Table of contents

Introduction

Evidence-based methods, cheapest first

This guide is ordered by how hard a method is to actually implement, not by how impressive its effect size is. A technique with strong evidence still gets demoted if it requires fighting entrenched habits or heavy upfront investment. Methods that interlock into a single loop are grouped in the same layer.

Suggested pace: Layer 0 can be finished today. Give Layer 1 two weeks to stabilize. Only then consider Layer 2. Layer 3 is optional — skip it when its preconditions aren’t met.

At a glance

Layer 0: Subtract First

This layer asks you to stop doing a few things and adjust two external conditions. The investment is near zero, which makes it the highest-return section in the guide.

0.1 Stop rereading and highlighting

Rereading produces an illusion of fluency. Text becomes familiar, the brain misreads familiarity as mastery, but the familiarity lives at the perceptual level and never becomes a retrievable memory.

Highlighting is a low-cognitive-load physical action. It creates the impression that the information has been processed — the work has been outsourced to the marker. In Fowler and Barker’s classic experiment, students who highlighted showed no significant overall advantage over those who didn’t. Peterson’s later work suggests underlining can actively hurt performance on inference questions, because attention is pulled toward isolated sentences and away from the logical connections between them.

Dunlosky and colleagues assessed 10 widely used study techniques. Five were rated low utility: rereading, highlighting, summarization, the keyword mnemonic, and imagery use for text learning.

These techniques are not negative — they simply have poor returns. Their real cost is displacing time that could have gone to retrieval practice.

Replacement: read the first pass normally, but without a marker. At the end of each section, close the book and say what it was about in one sentence. Put the reclaimed time into Layer 1.

Dunlosky, J., Rawson, K. A., Marsh, E. J., Nathan, M. J., & Willingham, D. T. (2013). Improving students’ learning with effective learning techniques. Psychological Science in the Public Interest, 14(1), 4–58. doi:10.1177/1529100612453266 · free PDF Fowler, R. L., & Barker, A. S. (1974). Effectiveness of highlighting for retention of text material. Journal of Applied Psychology, 59(3), 358–364. doi:10.1037/h0036750 Peterson, S. E. (1991). The cognitive functions of underlining as a study technique. Reading Research and Instruction, 31(2), 49–56. doi:10.1080/19388079209558078

0.2 Drop “learning styles”

The visual/auditory/kinesthetic learner taxonomy has no credible empirical support. Pashler, McDaniel, Rohrer, and Bjork reviewed the literature and found that the meshing hypothesis — that matching instructional format to a learner’s stated preference improves outcomes — has almost no support from adequately designed studies.

What actually matters is the nature of the material: geography benefits from maps, pronunciation from audio, physical technique from video. This is a property of the content, not of the learner.

The salvageable idea is dual coding — pairing text with visuals so the same information leaves a trace in two channels. Note that this is “use both,” not “pick the one that matches you.”

Pashler, H., McDaniel, M., Rohrer, D., & Bjork, R. (2008). Learning styles: Concepts and evidence. Psychological Science in the Public Interest, 9(3), 105–119. doi:10.1111/j.1539-6053.2009.01038.x Paivio, A. (1986). Mental Representations: A Dual Coding Approach. Oxford University Press. — origin of dual coding theory

0.3 Don’t trade sleep for study time

Memory consolidation happens largely during sleep; slow-wave sleep in particular supports the transfer of declarative memory from hippocampus to neocortex. Daytime encoding only stages the information — the write happens overnight.

Two extra hours gained by cutting sleep are frequently net negative: you lose that night’s consolidation and the next day’s encoding capacity and attention.

Practical form: schedule the hardest material for the 1–2 hours before bed, then sleep normally. Don’t fill that window with a phone.

Rasch, B., & Born, J. (2013). About sleep’s role in memory. Physiological Reviews, 93(2), 681–766. doi:10.1152/physrev.00032.2012 Diekelmann, S., & Born, J. (2010). The memory function of sleep. Nature Reviews Neuroscience, 11(2), 114–126. doi:10.1038/nrn2762

0.4 Single-task

Checking messages mid-study degrades encoding quality, not just elapsed time. The recovery cost of task switching is consistently larger than people estimate.

The evidence here is weaker than for retrieval practice — much of it is correlational. But since the implementation cost is essentially zero (put the phone in another room), it stays in the priority tier.

Rosen, L. D., Carrier, L. M., & Cheever, N. A. (2013). Facebook and texting made me do it: Media-induced task-switching while studying. Computers in Human Behavior, 29(3), 948–958. doi:10.1016/j.chb.2012.12.001 Monsell, S. (2003). Task switching. Trends in Cognitive Sciences, 7(3), 134–140. doi:10.1016/S1364-6613(03)00028-7

Layer 1: The Core Loop — Retrieve → Feedback → Space

This is the heart of the system. If time is limited and you do nothing else, this layer captures most of the available gain.

The three components must be used together; any one alone is discounted.

        ┌────────────────────┐
        │  1. Retrieve       │  Close the book. Pull it out.
        └─────────┬──────────┘
                  ↓
        ┌────────────────────┐
        │  2. Feedback       │  Check. Correct the errors.
        └─────────┬──────────┘
                  ↓
        ┌────────────────────┐
        │  3. Space          │  Return just before forgetting.
        └─────────┬──────────┘
                  ↓
              (back to 1)

1.0 Entry point: novices should study worked examples first

Worked example effect: for genuine beginners in a domain, studying complete solved problems outperforms attempting problems directly. The reason is that a novice’s working memory is consumed by search strategy, leaving no capacity to form schemas.

Expertise reversal effect: as knowledge accumulates, this advantage reverses. For learners with a foundation, direct practice beats studying examples.

Rule

• Unfamiliar domain → study 2–3 complete examples, then move to “completion problems” (first steps given, you finish).
• Familiar domain → solve directly, check afterward.

If you’re learning to cook, following a full recipe three times beats improvising from the start. Once you have a dozen dishes down, the recipe starts slowing you.

Sweller, J., & Cooper, G. A. (1985). The use of worked examples as a substitute for problem solving in learning algebra. Cognition and Instruction, 2(1), 59–89. doi:10.1207/s1532690xci0201_3 Kalyuga, S., Ayres, P., Chandler, P., & Sweller, J. (2003). The expertise reversal effect. Educational Psychologist, 38(1), 23–31. doi:10.1207/S15326985EP3801_4

1.1 Retrieval practice

The single best-evidenced intervention in the field. The principle: don’t push information in; force the brain to pull it out.

Each successful retrieval reshapes and strengthens the accessibility of that memory. Retrieval is itself a learning event, and a more effective one than re-exposure.

Evidence

• Karpicke & Roediger (2008, Science): foreign-language word-pair learning. One week later, recall was roughly 80% for the group that kept retrieving versus 36% for the group whose retrieval was dropped.
• Roediger & Karpicke (2006): with prose passages, one week later 61% (repeated testing) versus 40% (repeated study).
• Meta-analysis: Adesope, Trevisan, & Sundararajan (2017) pooled 272 experiments and found practice testing outperformed restudy at g ≈ 0.61. Rowland (2014) reports g ≈ 0.50.

Karpicke and Roediger used word pairs — pure associative memory. Effect sizes shrink when transferred to complex concepts, real classrooms, and long retention intervals. Treat g ≈ 0.5 as the realistic figure.

Practice

• Close the book. Reproduce the core concepts and structure on blank paper.
• When you’re stuck, sit with it for 1–2 minutes before looking. That uncomfortable interval is the part that works.
• Retrieve at a specific grain. “What happens to depth of field when I open the aperture, and why” beats “review photography.”

Karpicke, J. D., & Roediger, H. L. (2008). The critical importance of retrieval for learning. Science, 319(5865), 966–968. doi:10.1126/science.1152408 Roediger, H. L., & Karpicke, J. D. (2006). Test-enhanced learning: Taking memory tests improves long-term retention. Psychological Science, 17(3), 249–255. doi:10.1111/j.1467-9280.2006.01693.x Adesope, O. O., Trevisan, D. A., & Sundararajan, N. (2017). Rethinking the use of tests: A meta-analysis of practice testing. Review of Educational Research, 87(3), 659–701. doi:10.3102/0034654316689306 Rowland, C. A. (2014). The effect of testing versus restudy on retention: A meta-analytic review of the testing effect. Psychological Bulletin, 140(6), 1432–1463. doi:10.1037/a0037559

1.2 Immediate feedback

The benefit of retrieval practice depends heavily on whether a correct answer follows. Self-testing without feedback can fixate errors if you keep retrieving the wrong content — this is the standard way “quizzing myself” fails.

Butler and Roediger showed that providing feedback after a test substantially amplifies the testing effect, particularly by correcting high-confidence errors.

Rules

• After every recall attempt, check against the source. Non-negotiable.
• Flag two categories: answers you got wrong, and answers you got right but weren’t sure about. The second category is usually where you lose points next time.
• Feedback may be delayed by minutes (delayed feedback is sometimes better), but it cannot be absent.

Butler, A. C., & Roediger, H. L. (2008). Feedback enhances the positive effects and reduces the negative effects of multiple-choice testing. Memory & Cognition, 36(3), 604–616. doi:10.3758/MC.36.3.604 Butler, A. C., Karpicke, J. D., & Roediger, H. L. (2008). Correcting a metacognitive error: Feedback increases retention of low-confidence correct responses. Journal of Experimental Psychology: Learning, Memory, and Cognition, 34(4), 918–928. doi:10.1037/0278-7393.34.4.918

1.3 Spaced repetition

Rather than grinding five hours in one day, spread five one-hour sessions across two weeks.

The spacing effect: retrieval performed just as a memory becomes hard to access produces the strongest consolidation. Cepeda et al.’s meta-analysis of 254 studies, and Donovan and Radosevich’s earlier meta-analysis, put the effect size around d ≈ 0.4–0.5. This is robust and replicable.

How long a gap? Cepeda et al. (2008) give a usable rule: the optimal interval is roughly 10–20% of the target retention interval.

• Exam in one month → gaps of about 3–7 days.
• Need it a year from now → gaps stretching to 1–2 months.

The commonly cited 1-3-7-30 schedule is reasonable for an exam within the month. For long-term retention it is too dense past day 30 — keep extending rather than holding the interval constant.

Tools: Anki and other SRS software schedule this automatically. Building the cards is the upfront cost (and the reason for the three-star difficulty rating), but once running, the scheduling is free.

The corresponding failure mode — massed practice / cramming: excellent on an immediate test, then a cliff-edge drop on a delayed test one week later. It bypasses long-term consolidation, processing depth collapses under high-volume repetition, and the illusion of fluency is at its strongest. Cramming can rescue a deadline, but it cannot substitute for spacing — only sit on top of it.

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. doi:10.1037/0033-2909.132.3.354 Cepeda, N. J., Vul, E., Rohrer, D., Wixted, J. T., & Pashler, H. (2008). Spacing effects in learning: A temporal ridgeline of optimal retention. Psychological Science, 19(11), 1095–1102. doi:10.1111/j.1467-9280.2008.02209.x Donovan, J. J., & Radosevich, D. J. (1999). A meta-analytic review of the distribution of practice effect. Journal of Applied Psychology, 84(5), 795–805. doi:10.1037/0021-9010.84.5.795 Anki — free SRS software; uses FSRS as its default scheduler since version 23.10

Layer 2: Understanding — Interrogation and Self-Explanation

Layer 1 handles retention. Layer 2 handles comprehension. Difficulty rises sharply here, because this layer demands sustained active cognitive effort rather than a change of procedure.

The two methods are the light and heavy versions of the same thing, and pair well: elaborative interrogation runs while you read (low barrier, high frequency); the Feynman technique runs after (high barrier, low frequency).

2.1 Elaborative interrogation

As you meet new material, keep asking: Why is this true? Why this way rather than another?

This hooks new information onto existing background knowledge (schema), weaving isolated facts into a network. Dunlosky et al. rate elaborative interrogation as moderate utility. It helps with complex logical relationships, but its effectiveness depends strongly on the background knowledge you already have — with an empty background, there’s nothing to ask against.

Practice: when you hit a conclusion, derive its causes and consequences and connect it to something you already know. Learning photography, for instance:

1. Why does a smaller f-number let in more light?
2. Why does a slow shutter blur a handheld shot, but a slow shutter is exactly what you want for a waterfall?
3. Why does raising ISO let you shoot in the dark, at the cost of noise?

Answer all three and you’ll find they’re three faces of one question: how do you get the right quantity of light onto the sensor in the available time? That convergence is the network forming.

Dunlosky, J., et al. (2013). op. cit. — elaborative interrogation rated moderate utility. doi:10.1177/1529100612453266 Pressley, M., McDaniel, M. A., Turnure, J. E., Wood, E., & Ahmad, M. (1987). Generation and precision of elaboration effects on intentional and incidental learning. Journal of Experimental Psychology: Learning, Memory, and Cognition, 13(2), 291–300. doi:10.1037/0278-7393.13.2.291

2.2 The Feynman technique / self-explanation

Test the depth of your understanding through simplification and analogy. If you can’t explain it simply, you don’t understand it — and the act of explaining exposes the gaps.

Bisra et al.’s meta-analysis puts self-explanation at g ≈ 0.55. Dunlosky et al. likewise rate it moderate utility.

Implementation difficulty is high: it takes 10–20 minutes per concept, needs a quiet setting, and the process is unpleasant because it repeatedly reveals that you don’t actually understand. This is the step people abandon first.

Steps

1. Pick a concept.
2. Explain it as if to a ten-year-old.

   “Think of the camera as a bucket catching rain. The aperture is how wide the bucket’s mouth is; the shutter is how long you leave it out in the rain. Open the mouth wider, or leave it out longer, and you catch more water. ISO is how sensitive you are to the water — turn it up when there isn’t much rain, but turn it up too far and you start counting the splashes as real drops. That’s noise.”

3. Wherever you stall is a blind spot. Go straight back and reread that section.
4. Simplify the language further; find a better analogy.

Note that every good analogy also fails somewhere (the rain bucket says nothing about depth of field). Locating where the analogy breaks is itself a high-quality comprehension check.

Handoff to Layer 1: the Feynman technique is a high-quality retrieval attempt. When you finish, turn every point where you stalled into a flashcard and feed it directly into the Layer 1 loop.

Bisra, K., Liu, Q., Nesbit, J. C., Salimi, F., & Winne, P. H. (2018). Inducing self-explanation: A meta-analysis. Educational Psychology Review, 30(3), 703–725. doi:10.1007/s10648-018-9434-x Chi, M. T. H., de Leeuw, N., Chiu, M.-H., & LaVancher, C. (1994). Eliciting self-explanations improves understanding. Cognitive Science, 18(3), 439–477. doi:10.1207/s15516709cog1803_3

Layer 3: Interleaving

The hardest method here to implement, the most condition-dependent, and the easiest to get wrong. Consider it only once Layers 1 and 2 are running steadily.

Precondition: mix things that are genuinely confusable

The benefit of interleaving comes from discriminative contrast between similar, confusable material — the brain is forced to first decide which category is this, and which method applies?

Mixing photography, French vocabulary, and personal finance is not interleaving; it’s task switching. Those three can’t be confused with each other, so you pay the switching cost and collect none of the discrimination benefit.

Correct form

• Math: shuffle volume problems for spheres, cones, and wedges instead of doing a page of spheres.
• Photography: shuffle scenarios asking should I change aperture, shutter, or ISO here? instead of drilling aperture all day.
• Language: mix a handful of easily confused tenses or prepositions rather than one construction per page.
• Biology: contrast paired confusables — arteries vs. veins, mitosis vs. meiosis.

The test is simple: if you know which method to use without reading the problem, this practice set has no discriminative value.

Effect and boundaries

Rohrer and Taylor found that the interleaved group had a higher error rate and found practice harder, yet scored 63% versus 20% for the blocked group on a test one week later. Rohrer et al. (2020) replicated the effect in real math classrooms (roughly 61% vs. 38%).

Brunmair and Richter’s meta-analysis (59 studies, 238 effect sizes) found an overall effect of g = 0.42 with very large variation by material: strongest for paintings and other visual category learning (g = 0.67), small for mathematical tasks (g = 0.34), ambiguous and non-significant for expository texts, and negative for word learning (g = −0.39), where blocking actually won. Do not apply this method unconditionally.

Psychological cost: practice feels distinctly worse and accuracy drops, which makes it easy to conclude the method isn’t working and quit. That’s the main reason for the five-star difficulty rating.

Rohrer, D., & Taylor, K. (2007). The shuffling of mathematics problems improves learning. Instructional Science, 35(6), 481–498. doi:10.1007/s11251-007-9015-8 Rohrer, D., Dedrick, R. F., Hartwig, M. K., & Cheung, C.-N. (2020). A randomized controlled trial of interleaved mathematics practice. Journal of Educational Psychology, 112(1), 40–52. doi:10.1037/edu0000367 Brunmair, M., & Richter, T. (2019). Similarity matters: A meta-analysis of interleaved learning and its moderators. Psychological Bulletin, 145(11), 1029–1052. doi:10.1037/bul0000209 · author PDF

The Underlying Principle: Desirable Difficulty

Bjork’s desirable difficulties framework: moderate difficulty during learning triggers deeper encoding and produces stronger long-term retention and transfer. Every uncomfortable element in Layers 1 through 3 traces back to this mechanism.

The boundary: difficulty is only desirable when it is surmountable. Imposing difficulty beyond a novice’s capacity produces failure, abandonment, and mis-encoding. This is precisely why “worked examples first” (§1.0) has to sit ahead of retrieval practice.

The test: if you can recover the answer after struggling for 1–2 minutes, the difficulty is productive. If you’re completely blank with no partial trace, encoding never finished — go back and relearn rather than grinding.

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). MIT Press. Bjork, E. L., & Bjork, R. A. (2011). Making things hard on yourself, but in a good way: Creating desirable difficulties to enhance learning. In Psychology and the Real World. Worth Publishers. — Bjork Learning and Forgetting Lab

Metacognitive Calibration

The root of the illusion of fluency is that people are systematically overconfident about whether they’ve learned something.

Judgment accuracy is itself a trainable skill: before each self-test, predict whether you’ll get the item right; after answering, check the prediction against the outcome. After a few rounds you’ll see which categories of material you most reliably overestimate — usually exactly where the effort should go.

Working signal: recognizing something is not being able to state it. Anything that only reaches recognition hasn’t been learned.

Koriat, A., & Bjork, R. A. (2005). Illusions of competence in monitoring one’s knowledge during study. Journal of Experimental Psychology: Learning, Memory, and Cognition, 31(2), 187–194. doi:10.1037/0278-7393.31.2.187 Dunning, D., Johnson, K., Ehrlinger, J., & Kruger, J. (2003). Why people fail to recognize their own incompetence. Current Directions in Psychological Science, 12(3), 83–87. doi:10.1111/1467-8721.01235

Failure Modes and Their Replacements

The Workflow

Illustrated with “learning photography.” Substitute any body of knowledge that has internal structure.

Day 0 — the study session

Day 1 onward — consolidation

If you need the material a year out, keep extending the interval past day 30 to every 2–3 months rather than holding at 30 days.

A note on all the effect sizes above

Nearly all of these numbers come from laboratory settings, short retention intervals, and relatively simple materials. Moved into real classrooms, across months, against complex concepts, effects shrink broadly — and some methods (interleaving especially) become unstable in particular domains.

The ordering of these methods is reliable. The specific percentages are not.

Full reference list

Meta-analyses and reviews

• Adesope, O. O., Trevisan, D. A., & Sundararajan, N. (2017). Rethinking the use of tests. Review of Educational Research, 87(3), 659–701. doi
• Bisra, K., et al. (2018). Inducing self-explanation: A meta-analysis. Educational Psychology Review, 30(3), 703–725. doi
• Brunmair, M., & Richter, T. (2019). Similarity matters. Psychological Bulletin, 145(11), 1029–1052. doi
• Cepeda, N. J., et al. (2006). Distributed practice in verbal recall tasks. Psychological Bulletin, 132(3), 354–380. doi
• Donovan, J. J., & Radosevich, D. J. (1999). A meta-analytic review of the distribution of practice effect. Journal of Applied Psychology, 84(5), 795–805. doi
• Dunlosky, J., et al. (2013). Improving students’ learning with effective learning techniques. PSPI, 14(1), 4–58. doi
• Pashler, H., et al. (2008). Learning styles: Concepts and evidence. PSPI, 9(3), 105–119. doi
• Rasch, B., & Born, J. (2013). About sleep’s role in memory. Physiological Reviews, 93(2), 681–766. doi
• Rowland, C. A. (2014). The effect of testing versus restudy on retention. Psychological Bulletin, 140(6), 1432–1463. doi

Primary studies

• Butler, A. C., & Roediger, H. L. (2008). Memory & Cognition, 36(3), 604–616. doi
• Cepeda, N. J., et al. (2008). Spacing effects in learning. Psychological Science, 19(11), 1095–1102. doi
• Chi, M. T. H., et al. (1994). Cognitive Science, 18(3), 439–477. doi
• Diekelmann, S., & Born, J. (2010). Nature Reviews Neuroscience, 11(2), 114–126. doi
• Fowler, R. L., & Barker, A. S. (1974). Journal of Applied Psychology, 59(3), 358–364. doi
• Kalyuga, S., et al. (2003). The expertise reversal effect. Educational Psychologist, 38(1), 23–31. doi
• Karpicke, J. D., & Roediger, H. L. (2008). Science, 319(5865), 966–968. doi
• Koriat, A., & Bjork, R. A. (2005). JEP:LMC, 31(2), 187–194. doi
• Peterson, S. E. (1991). Reading Research and Instruction, 31(2), 49–56. doi
• Roediger, H. L., & Karpicke, J. D. (2006). Psychological Science, 17(3), 249–255. doi
• Rohrer, D., & Taylor, K. (2007). Instructional Science, 35(6), 481–498. doi
• Rohrer, D., et al. (2020). Journal of Educational Psychology, 112(1), 40–52. doi
• Sweller, J., & Cooper, G. A. (1985). Cognition and Instruction, 2(1), 59–89. doi