Which Study Methods Work: What Research Says About Self-Testing and Spacing

Self-testing and spaced study have the strongest research support among common methods, while rereading and highlighting have a weaker record. Effect sizes vary by material, delay and feedback, and some real-course results are modest. Students also tend to misjudge which methods help.

Practicing recall and spreading study sessions over time have the strongest research backing among common study methods. Rereading and highlighting, the habits students most often report, have a weaker record. That is the central conclusion of a large 2013 review of ten learning techniques, and several meta-analyses since then point the same way.

The direction of the evidence is clearer than its size. Effects shrink in some real courses, and results depend on the material, the delay before the test and whether feedback is given. The stakes are practical, because students tend to judge a method by how smooth it feels, and the research suggests that feeling can mislead.

What the broadest review concluded

The most cited overview is a 2013 monograph by John Dunlosky and colleagues in Psychological Science in the Public Interest. It evaluated ten techniques that students commonly use or are advised to use. For each one, the authors asked whether its benefits hold across four categories: learning conditions, student characteristics, materials and criterion tasks.

• Elaborative interrogation
• Self-explanation
• Summarization
• Highlighting and underlining
• Keyword mnemonics
• Imagery use for text learning
• Rereading
• Practice testing
• Distributed practice
• Interleaved practice

According to Dunlosky's own account in American Educator, practice testing and distributed practice earned the top ratings. His reasoning was that they help students of different ages, work across a wide range of materials and raise achievement. The review also found that most students report rereading and highlighting, yet these do not reliably improve performance, so the authors recommended alternatives such as practice testing in place of rereading.

The monograph also reports that prompting students to answer "why" questions, as in elaborative interrogation and self-explanation, can facilitate learning. A table row in the same document rates imagery use for text learning as low. A blog summary places interleaving and self-explanation in a moderate tier, but that comes from a secondary page and has not been checked against the original table.

The review has limits. It is a narrative review that rates techniques, not a meta-analysis that pools effect sizes, and it predates the larger classroom meta-analyses discussed below. Its ratings are best read as a well-informed starting point rather than a final ranking.

Retrieval practice: a consistent benefit, with a size that depends on the comparison

How large the effect is

Several independent meta-analyses find that testing beats restudying. A 2014 meta-analysis by Rowland in Psychological Bulletin found a reliable medium effect of testing versus restudy, g = 0.50, according to a review of retrieval practice in the health professions. A 2018 meta-analysis by Pan and Rickard found a medium effect on transfer of test-enhanced learning, d = 0.40, as reported in the same review.

A larger 2017 meta-analysis by Adesope, Trevisan and Sundararajan, in Review of Educational Research, covered 15,427 participants across 272 comparisons and reported an average effect of g+ = 0.70, according to a guide from a retrieval-practice advocacy site. That headline figure is inflated by comparisons against weak control conditions, and the more conservative number is the one measured against restudying. The practical lesson is to read the comparison group before the effect size, since retrieval looks larger against doing nothing than against rereading.

What happens in real classrooms

Much early retrieval research used word lists and short texts in laboratories, so the classroom question matters. A 2019 review in Frontiers in Education found clear testing effects in classrooms when retrieval practice was compared with weak strategies such as rereading, highlighting or a brief lecture exposure, or with no activity at all. A 2021 meta-analysis by Yang and colleagues in Psychological Bulletin, titled "Testing (quizzing) boosts classroom learning," is described by a commercial summary as covering 222 studies and 48,478 students with a medium effect of g = 0.499. Those figures come from a secondary page and have not been confirmed against the paper itself.

Evidence for narrower settings is thinner. A 2017 meta-analysis of the testing effect in psychology classrooms, by Schwieren and colleagues, concluded that retrieval practice can foster learning, but it rested on only 19 studies, a relatively small number.

Format and feedback

According to a Learning Scientists blog post summarizing the 2017 Adesope analysis, hybrid practice tests that combined multiple-choice and cued recall worked best. The same post says retrieval with feedback was only slightly more effective than retrieval without it, and that the largest effects appeared among secondary school students. Other secondary summaries of the Yang analysis suggest a bigger feedback gap, so how much feedback adds remains unresolved until the primary papers are compared.

Spaced practice: a robust direction, a modest size in real courses

The evidence base for spacing is large. A 2006 review by Nicholas Cepeda, Harold Pashler and colleagues in Psychological Bulletin pooled 839 assessments of distributed practice from 317 experiments in 184 articles. It found that the gap between study sessions and the delay before the final test act together: the spacing that produced the best retention grew longer as the retention interval grew longer. The review covered verbal recall tasks, which are mostly laboratory studies.

A meta-analysis of applied research on classroom learning, published in PMC, found a moderate advantage for distributed over massed practice, d = 0.54, with a 95% confidence interval of 0.31 to 0.77. Its authors acknowledge that the extensive spacing literature comes predominantly from laboratories, often with decontextualized material. The same page cites a larger d = 0.85 from a synthesis of learning strategies, but that synthesis draws largely on articles from an earlier review of the same literature, so it is not independent confirmation.

Combining spacing with retrieval

A meta-analysis of 29 studies reported a strong benefit of spaced over massed retrieval practice, g = 0.74. The retrieved copy of that paper did not show its authors or date, and it may overlap with the other literature discussed here, so the figure warrants caution. It supports the idea that pairing the two methods is sensible, with moderate rather than high confidence.

A cautionary result from actual courses

A 2024 paper in the International Journal of STEM Education pooled single-paper meta-analyses of spaced retrieval practice in nine introductory STEM courses. It estimated a gain of 2.06 percentage points (95% CI 0.16 to 3.97), which was statistically significant across the nine courses but small in absolute terms. The authors also discuss a non-significant effect across eight courses and earlier findings of boundary conditions: classroom settings, no feedback and multiple-choice retrieval. Relative effect sizes in the medium range and an absolute gain of about 2 points can both be true, because they measure different things.

Interleaving: useful for some material, not all

Interleaving mixes categories or problem types instead of practicing one kind at a time. Dunlosky's account in American Educator cites improvements over massed practice in fourth-graders' math, engineering students diagnosing system failures, college students learning artists' styles and medical students interpreting ECGs. That is a varied set of domains, though it is a list of examples rather than a pooled estimate.

A 2019 meta-analysis by Brunmair and Richter in Psychological Bulletin found a pooled advantage of interleaving over blocking of g = 0.42, with a 95% confidence interval of 0.34 to 0.50. A database tally describes the analysis as 59 studies, 8,466 participants and 238 effects. For mathematical tasks specifically, the effect was g = 0.34 (95% CI 0.11 to 0.57), according to a copy of the paper hosted on ResearchGate.

Material appears to matter. A Loughborough University blog summarizing the results says interleaving might not benefit the learning of words but could help with classifying paintings. Another research team argues that the negative conclusion about verbal materials may be premature once the included studies are examined more closely. In mathematics, primary studies ranged from strongly negative to positive, so the average hides wide variation.

One proposed mechanism is discriminative contrast: seeing categories side by side highlights what distinguishes them. A secondary summary describes this idea, and it should be treated as a theory, not a settled finding.

The numbers side by side

Why good methods feel bad, and what learning styles add

The gap between feeling and learning

A 2019 randomized experiment by Louis Deslauriers and colleagues in PNAS used identical materials in Harvard physics classes. Students in the active classroom learned more but felt they had learned less. The authors say students read the extra cognitive effort as a sign of poorer learning, and the lead author told ScienceDaily that actual learning and the feeling of learning were strongly anticorrelated.

The finding has limits. It comes from one institution with an elite student population, it concerns classroom instruction rather than individual studying, and sample size and effect sizes were not retrieved for this article. It is consistent with a plausible explanation for why rereading remains popular, but it does not by itself show that students pick study methods this way.

Learning styles

Preferences for how to learn exist, but the claim that teaching to them improves learning lacks support. A 2008 review by Harold Pashler and colleagues, commissioned by the Association for Psychological Science, set out the test a valid study needs: a crossover interaction, in which each group does best with its own matched method. Their conclusion, as quoted by a secondary page:

at present, there is no adequate evidence base to justify incorporating learning-styles assessments into general educational practice

A later empirical test by Rogowsky, Calhoun and Tallal, published in 2014 in the Journal of Educational Psychology, found no statistical support for the matching hypothesis. It examined verbal comprehension aptitude and mode of instruction, comparing audiobooks with e-text.

A recent meta-analysis in PMC complicates the story. It found an overall positive effect of matching instruction to modality preferences, g = 0.32 (95% CI 0.07 to 0.57). Yet only 26% of learning outcome measures showed matched-instruction benefits for at least two styles, which is the crossover pattern the hypothesis requires. That second figure favors the skeptical view, and the paper's authors and date were not retrieved, so it deserves a full reading before anyone calls the question closed.

Belief in learning styles persists. A 2020 systematic review in Frontiers in Education found high self-reported belief among educators in matching instruction to styles, with no evidence that the belief is supported.

What the evidence adds up to

The phrase "the best method" does not hold up well against the research. What the evidence supports is a small set of strategies, namely retrieval, spacing and, for some materials, interleaving, that tend to beat common habits such as rereading and highlighting. Confidence is high that self-testing beats rereading and that spacing beats massing in direction, and moderate on how large the benefit is in real courses.

The harder problem may be behavioral. If effective methods feel worse while they work, students will keep choosing the comfortable ones unless something corrects the impression. The next useful evidence will come from full-course trials that report absolute gains and long-term retention, and from preregistered replications that show how much of the laboratory advantage survives outside the lab.

Frequently asked questions

Which study techniques have the strongest evidence?

In the 2013 Dunlosky review, practice testing and distributed practice earned the top ratings because they help students of different ages, work across many materials and raise achievement. Meta-analyses since then report medium effects for retrieval practice and for spacing. Effect sizes vary with the comparison and the setting.

Is rereading a bad way to study?

The Dunlosky review found that rereading and highlighting, the methods most students report using, do not consistently boost performance. Classroom studies also find retrieval practice beating rereading. The evidence here comes from reviews, not from primary rereading studies retrieved for this article.

Does spaced practice work in real classes?

The direction is favorable. A classroom meta-analysis found a moderate effect of d = 0.54 for distributed over massed practice. A study of nine introductory STEM courses found a gain of about 2 percentage points for spaced retrieval, which is small in absolute terms.

Does interleaving help with every subject?

No. A 2019 meta-analysis found an overall pooled advantage of g = 0.42, but effects depend on the material. Results were weaker or possibly negative for some verbal content and varied widely among mathematics studies.

Do learning styles improve learning?

A 2008 review found no adequate evidence base for using learning-styles assessments in general educational practice, and a 2014 test found no support for matching. A recent meta-analysis reported a small positive matching effect (g = 0.32), but only 26% of outcome measures showed the crossover pattern that the hypothesis requires.

Sources

1. Improving Students' Learning With Effective Learning Techniques , Psychological Science in the Public Interest (via Semantic Scholar)
2. Improving Students' Learning With Effective Learning Techniques (PSPI 14(1), 4–58) , Psychological Science in the Public Interest (copy hosted by whz.de)
3. Strengthening the Student Toolbox , American Educator (AFT)
4. The Use of Retrieval Practice in the Health Professions: A State-of-the-Art Review , PMC
5. Effect Sizes and Meta-Analyses , Retrieval Practice (retrievalpractice.org)
6. New Meta-analysis of 217 Retrieval Practice Studies , The Learning Scientists (2017-02-09)
7. Retrieval Practice in Classroom Settings: A Review of Applied Research , Frontiers in Education
8. The Testing Effect in the Psychology Classroom: A Meta-Analytic Perspective , SAGE Journals
9. Retrieval practice and the testing effect , MicroPlay Games
10. Cepeda, Pashler, Vul, Wixted, and Rohrer, 2006 (Distributed practice in verbal recall tasks) , York University
11. The Distributed Practice Effect on Classroom Learning: A Meta-Analytic Review of Applied Research , PMC
12. A Meta-Analytic Review of the Benefit of Spacing , Educational Psychology Review (PDF via lscp.net)
13. Single-paper meta-analyses of the effects of spaced retrieval practice in nine introductory STEM courses: is the glass half full or half empty? , International Journal of STEM Education (Springer) (2024-01-01)
14. Similarity Matters: A Meta-Analysis of Interleaved Learning and Its Moderators , ResearchGate (Brunmair & Richter, Psychological Bulletin) (2019-01-01)
15. A systematic review of interleaving as a concept learning strategy , Review of Education (Wiley) (2021-01-01)
16. Using Interleaving to Promote Inductive Learning in Educational Contexts: Promises and Challenges , Hogrefe
17. Spaced and interleaved practice , Loughborough University blog (2024-04-09)
18. Visible Learning - Interleaved practice Details , Visible Learning MetaX
19. Interleaving: when mixing topics helps you learn, and when it does not , Psychology World
20. Measuring actual learning versus feeling of learning in response to being actively engaged in the classroom , PNAS (via Semantic Scholar) (2019-01-01)
21. More learning in 'active learning' classrooms, but students don't know it , ScienceDaily (2019-09-06)
22. From Styles to Science: Debunking the Learning Styles Myth and Embracing an Evidence-Based Framework for Learning , Global Council for Behavioral Science
23. Learning Styles: Out of Fashion , American Psychological Association
24. Is it really a neuromyth? A meta-analysis of the learning styles matching hypothesis , PMC
25. Learning Styles Neuromyth, and Systematic Review , Frontiers in Education (2020-01-01)

This article was drafted with the help of AI and reviewed by Jaroha editors before publication.