Prompting learners to self-explain improves understanding and problem solving on immediate tests by a moderate average amount, with little evidence yet on delayed or classroom gains.¶

Claim · All claims
Evidence · 4 studies · 2 quant-synthesis r?, 1 causal r1, 1 associational r1 · q2–q4 · i2 medium

Generating explanations for oneself while studying worked examples or solving problems helps learners bridge gaps in their knowledge, integrate new information with prior schemas, and monitor their own comprehension.

Subclaims¶

q2 i? Students who generate more self-explanations while studying worked examples perform significantly better on subsequent problem-solving tasks. → Chi et al. 1989 q3 i? Prompted self-explanation leads to moderate improvements in conceptual knowledge and procedural transfer in mathematics. → Rittle-Johnson 2017

q4 i2 Across 69 effect sizes from 64 reports, prompting learners to self-explain while studying or solving problems improved learning outcomes by a moderate average amount (g = .55), across task types, subject areas and levels of education. → Bisra et al. 2018

q3 i? In mathematics, prompted self-explanation gives a small-to-moderate immediate improvement in procedural knowledge, conceptual knowledge and procedural transfer, but evidence for gains in classroom settings or after a delay is much more limited. → Rittle-Johnson 2017

q2 i? Eighth-grade students prompted to self-explain each line of a text on the circulatory system gained more from pretest to posttest than students who read it twice, and those who explained most built the most accurate mental models. → Chi et al. 1994

Evidence¶

Primary evidence link: https://doi.org/10.1207/s15516709cog1302_1

Chi et al. 1989¶

Chi, M. T. H., Bassok, M., Lewis, M., Reimann, P., & Glaser, R. (1989). Self-explanations: How students study and use examples in learning to solve problems. Cognitive Science, 13(2), 145–182. https://doi.org/10.1207/s15516709cog1302_1

q2 · observational study of 10 students, no manipulated condition · i? · the abstract prints no effect size; the full text may · n=10 · associational · r1

The seminal study identifying the "self-explanation effect." High-achieving students were found to generate significantly more self-explanations (52 vs. 18) while studying physics examples, using them to connect steps to underlying principles and monitor their own understanding.

Rittle-Johnson 2017¶

Rittle-Johnson, B., Loehr, A. M., & Durkin, K. (2017). Promoting self-explanation to improve mathematics learning: A meta-analysis and instructional design principles. ZDM Mathematics Education, 49(4), 599–611. https://doi.org/10.1007/s11858-017-0834-z

q3 · meta-analysis · i? · no source text available to check; the entry prints no effect size · n=N/A · quant-synthesis · r?

A meta-analysis of studies in mathematics education. It found that prompted self-explanation consistently improves both conceptual and procedural knowledge and that prompts are most effective when they focus on conceptual "why" questions rather than procedural "what" questions.

Bisra et al. 2018¶

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

q4 · meta-analysis (random effects) · i2 · medium effect, g=.55 · n=69 effect sizes from 64 research reports · quant-synthesis · r?

A systematic search and meta-analysis of studies in which participants received self-explanation prompts while studying or solving problems, compared with learners who did not. The random-effects weighted mean effect was g = .55. The authors coded 20 moderators, including the type of task (solving problems, studying worked examples, studying text), subject area, level of education, type of prompt and treatment duration, and conclude that self-explanation prompts are a potentially powerful intervention across a range of instructional conditions. Read from the abstract only; moderator results are not reported here.

Chi et al. 1994¶

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

q2 · small controlled experiment, two groups · i? · no effect size in the abstract · n=24 students · causal · r1

Fourteen eighth-grade students were asked, with no extensive training, to self-explain after reading each line of an expository passage on the human circulatory system; ten control students read the same text twice without prompts. All were tested before and after. The prompted group gained more from pretest to posttest, and within it the students who generated many self-explanations learned with greater understanding than those who generated few. All high explainers reached the correct mental model of the circulatory system, while many unprompted students and low explainers did not.

Studies bearing on this claim¶

Does prompting learners to explain what they study or solve improve their understanding, and on which outcomes? This table gathers the studies behind this claim and the 16 narrower claims that bear on it: 15 distinct studies (5 causal, 4 review, 2 associational, 2 quant-synthesis, 1 theoretical, 1 qualitative), q2–q4, 1 of 15 with an effect size coded. One row per study, codes as each entry records them; no effect is averaged and no verdict is drawn. Built by scripts/synthesize_claim.py from eval/synthesis/; a row that looks wrong is fixed on the claim it comes from.

Discussion¶

The effectiveness of self-explanation can be limited by a learner's prior knowledge; if a learner lacks the necessary background to generate a correct explanation, they may reinforce misconceptions.

Merged from “Self-explanation improves learning” (self-explanation-improves-learning): Mechanism. Self-explanation is thought to work by closing gaps in mental models: when learners must state why a step follows or how an example relates to a principle, they generate inferences that passive study leaves implicit [+M]. This aligns with the broader account in Cognitive Load Theory — explanation prompts direct limited working-memory resources toward schema construction rather than surface features, in a manner related to how Chunking reduces working memory load [+M].

Moderators and boundary conditions. The effect is generally strongest for novices, who have the most gaps to close [+M]; for advanced learners, explanation prompts can become redundant or even disruptive, consistent with the Expertise Reversal Effect [~M]. Prompt quality also matters: open prompts ("explain this to yourself") can leave unprepared learners explaining superficial features [~M], while more structured or scaffolded prompts tend to yield more productive explanations [+M]. Because generating explanations adds effort, poorly timed or excessive prompting risks overload — the same failure mode described in Cognitive overload degrades learning [-M].

Relation to worked examples. Self-explanation prompts are frequently layered onto worked examples, where they prompt learners to justify each step rather than passively read it [+M]. The two strategies are complementary: examples reduce search during early learning, while explanation prompts ensure the example's underlying principles are actually extracted rather than copied — see Example–problem sequences reduce cognitive load and improve learning outcomes.

Relation to other generative strategies. Self-explanation belongs to a family of generative techniques — annotating, summarizing, questioning — that share a common logic: learning improves when learners actively construct connections rather than receive them [+M]. What distinguishes self-explanation is its focus on justifying steps and relating representations, which makes it especially well suited to worked examples, case-based learning, and step-by-step procedures where each step has an underlying rationale worth articulating.

Open questions. Four studies are now recorded above: two meta-analyses (Bisra et al. 2018, g = .55 across task types and subjects; Rittle-Johnson 2017 in mathematics), one small controlled experiment (Chi et al. 1994) and one observational study (Chi et al. 1989). Domain boundaries and the durability of gains over time are still thinly documented: Rittle-Johnson 2017 finds evidence after a delay or in classrooms much more limited, and the moderator results of Bisra et al. are not recorded here. Key questions for future enrichment: how self-explanation compares with other generative strategies of equal time-on-task, whether prompted explanation produces durable gains or only momentary engagement, and how prompt structure should be faded as expertise develops.

Merged from “Eliciting Self Explanations Improves Understanding” (eliciting-self-explanations-improves-understanding): Mechanism. Self-explanation prompts are usually understood through generative-learning logic: producing an explanation forces the learner to connect new information to what they already know, exposing gaps and repairing flawed mental models. This places self-explanation close to Activation and Analogical reasoning as integration-supporting techniques.

Costs and boundary conditions. Generating explanations consumes working-memory resources, so for novices or very high-element-interactivity material the prompt itself can add load rather than support integration [~M] — the same trade-off that governs Chunking and Cognitive load theory more broadly. When material is already well understood, explaining it can become redundant and yield little or no benefit, echoing the expertise-reversal pattern described in Expertise reversal effect [~M]. Prompt design matters: open-ended "explain this" prompts can produce shallow restatements, while prompts that target principles or underlying reasons are more likely to generate productive explanations [~M].

Design implications. Practitioners should (a) attach explanation prompts to specific steps or statements rather than asking for global summaries, (b) scaffold early prompts with sentence starters or principle cues and fade them as competence grows, and (c) avoid prompting explanation of material learners can already recite fluently — at that point the prompt is redundant effort with no integration payoff [~M]. Self-explanation pairs naturally with worked examples and with case-based learning, where prompts can direct attention to why a case or solution step unfolds as it does.

Open questions. The durability of self-explanation effects over delayed tests, their interaction with learner expertise, and the relative value of self-generated versus instructional explanations remain active areas of investigation. The four studies recorded above test immediate learning; none of them settles delayed retention or the interaction with expertise.

Merged from “Self Explanation Prompts Improve Learning” (self-explanation-prompts-improve-learning): Mechanism. Self-explanation prompts work by forcing learners to go beyond passive restatement: they map new information onto prior knowledge, expose gaps in understanding, and help learners infer the underlying principles behind worked steps or textual claims. This aligns with the broader generative-learning position that learning improves when learners actively construct connections rather than receive explanations — see Activation improves learning [+M] and Active learning improves exam performance [+S].

Boundary conditions. The benefit is strongest for novices, who otherwise skip self-explanation or do it shallowly. For more advanced learners, prompts can become redundant or interrupt fluent processing — the same expertise-reversal pattern documented for worked examples and formalized in expertise reversal [~M]. Prompt design also matters: open-ended "explain why" prompts can overload limited working memory if the material is high in element interactivity, in which case providing the key explanation (an "assisted" or scaffolded prompt) can outperform pure self-generation [~M]. This connects to Cognitive overload degrades learning [-S] and Chunking reduces working memory load [+S], and situates the strategy within cognitive load management.

Prompt placement and format. Prompts can be attached to worked steps, textual passages, or multimedia segments. Principle-based prompts ("what principle does this step illustrate?") and gap-targeted prompts tend to focus explanation on the deep structure of the material, whereas vague prompts ("explain this to yourself") risk eliciting paraphrase rather than genuine elaboration [~W]. Written explanations impose more load than oral or prompted-selection formats, so designers of annotating-style tasks should weigh response format against material complexity.

Open questions. The durability of gains on delayed tests, and how prompt type (principle-based vs. goal-based vs. gap-targeted) moderates outcomes, are not settled by the evidence recorded above.

Merged from “Self Explanation Prompts Improve Learning From Worked Examples” (self-explanation-prompts-improve-learning-from-worked-examples): Mechanism. Self-explanation prompts are thought to work by directing attention to the underlying principles and rationale of each solution step, helping learners build schemas that connect the example to domain knowledge — consistent with Cognitive Load Theory accounts of worked-example learning. Without prompts, learners often passively read examples and fail to engage in the inferential processing that produces transferable schemas; prompts convert example study from reading into an active, generative act, in line with Activation of relevant prior knowledge.

Moderators and boundary conditions. The benefit of self-explanation prompts is widely understood to interact with learner expertise: for novices, prompts scaffold productive processing, but for more advanced learners the same prompts can impose extraneous load or become redundant — the expertise reversal pattern described in Worked examples can become redundant or counterproductive for advanced learners. and formalized in the Expertise Reversal Effect. Prompt design also matters: open-ended "why" prompts can be demanding, and learners who generate poor or shallow explanations may gain little; the added processing can itself be a source of cognitive overload when examples are complex or working-memory resources are stretched. Designers should consider providing partial explanations, fading prompts over time, and pairing prompts with worked examples interleaved with practice rather than examples alone. Prompt timing is another open design variable: prompts attached to every step may disrupt the smooth reading of a solution, whereas prompts placed after complete solution steps tend to preserve the example's coherence while still eliciting explanation.

Open questions. Optimal prompt formats and the durability of gains across domains are not settled by the evidence recorded above, which concerns self-explanation in general rather than worked examples specifically.

Related Claims¶

• Questioning Strategies Improve Learning — the general claim this finding bears on, with the studies behind it

• Self-questioning improves comprehension — the general claim this finding bears on, with the studies behind it

• Constructive learning beats active and passive learning — the general claim this finding bears on, with the studies behind it

• Instructional guidance that helps novices can become redundant or counterproductive as expertise grows.

• Chunking reduces working memory load by grouping information into fewer, more meaningful units.
• Erroneous examples improve conceptual understanding by forcing comparison with correct models. — related
• Activation improves learning — self-explanation works by connecting new material to activated prior knowledge
• Learner highlighting of text gives a small memory benefit but no reliable comprehension benefit, and evidence on other forms of annotation is not yet recorded here — a related generative strategy that externalizes learner thinking
• Active learning improves exam performance — self-explanation is one generative mechanism within active learning
• Coherence principle: irrelevant material hurts learning — explanation prompts are most productive when attention is directed to relevant content
• Example–problem sequences reduce cognitive load and improve learning outcomes — self-explanation prompts are commonly layered onto worked examples in these sequences
• Analogical reasoning improves transfer — explaining how cases relate supports transfer
• Cognitive overload degrades learning — poorly timed explanation prompts can overload rather than support
• Expertise reversal effect — benefits shrink or reverse as learner expertise grows
• Worked examples reduce unnecessary search for novices. — self-explanation prompts are frequently layered onto worked examples to deepen their processing
• Worked examples can become redundant or counterproductive for advanced learners. — the same expertise-reversal moderation likely applies to explanation prompts
• Cognitive load reduction improves learning. — prompt design must avoid adding extraneous load that offsets generative benefits
• Elaborative Encoding Improves Retention — possibly the same claim (merge candidate)
• Embedded aids address the content-area reading problem by helping weak readers read textual material and helping effective readers transfer training across disciplines, via a running commentary explaining key concepts, terminology, and questions. — a narrower finding that bears on this claim
• Strategy Instruction Improves Learning — related
• Students with more controlled interaction patterns in iSTART-2 generated higher-quality self-explanations than students with more random patterns — a narrower finding that bears on this claim
• Example-based sequences outperform problem-only practice for novices, and fading support is argued to aid transfer — prompts can be faded alongside example–problem transitions
• Studying worked examples improves novices' algebra performance on problems with the same structure as the examples — the population where self-explanation prompts are most often studied
• Chunking reduces working memory load. — well-chunked examples keep self-explanation demands within capacity
• Cognitive load management — prompts add generative load, so total load must stay within capacity
• Teaching reflective questioning or using graphical representations improves procedural and conceptual knowledge across diverse populations — related
• Theory-guided five-step prompting of AI interaction improves open-ended explanation performance in construction education review, while unprompted AI access alone does not — a narrower finding that bears on this claim
• Immediate LLM explanations risk bypassing the generative processes required for durable understanding — related
• Generative AI makes knowledge mobilizable, converting external knowledge into immediately usable task output within seconds — related
• Faded steps were hypothesized to trigger specific self-explanation activities, tested via think-aloud protocols in Experiment 2 — related
• Self-explanation prompting is potentially effective across a range of instructional conditions — related
• Students who self-explain worked examples learn more effectively than those who do not, and the quality of self-explanations relates to explaining deep structure — possibly the same claim (merge candidate)
• Self-explanation effects hold on far-transfer but not near-transfer questions in a geometry classroom study — a narrower finding that bears on this claim
• Self-explanation improves abstract transfer of logical rules relative to no self-explanation — a narrower finding that bears on this claim
• Learning through the experience of solving problems can build both content knowledge and thinking strategies — related
• Assisting self-explanation prompts particularly foster conceptual knowledge, an effect mediated by self-explanations on the rationale of a principle — a narrower finding that bears on this claim
• The assisting self-explanation prompt effect: assisting prompts elicit high-quality self-explanations and deep understanding — a narrower finding that bears on this claim
• Providing worked examples of journal entries early (rather than delayed) yields better early journal quality and more conceptual knowledge — related
• Explaining one's own potentially incorrect ideas can reduce learning, while explaining known-correct content supports it — related
• Explanation prompts direct attention to targeted information and can reduce attention to other important information — related
• Self-explanation beats unguided studying for equal time, but instructional explanations, problem solving and retrieval practice can be equally or more effective — related
• Self-explanation promotes transfer of general patterns but can reduce recall of details and cause overgeneralization — a narrower finding that bears on this claim
• Both open and assisting self-explanation prompts foster procedural knowledge when learning with multiple representations, an effect mediated by self-explanations directed to domain principles — a narrower finding that bears on this claim
• Erroneous examples produce significantly higher delayed-test learning gains than problem solving for decimal learning — a narrower finding that bears on this claim
• Low prior knowledge students learned most conceptual understanding from worked-out examples, and least from questions plus arguments — related
• Explicit training in constraint-seeking questions showed only moderate, non-generalizing benefits, whereas prompting explanation supports better question asking — a narrower finding that bears on this claim
• Students who collected more peers' ideas wrote poorer explanations, while generating own redundant ideas related to better explanations — related
• Self-explaining worked examples improve declarative knowledge of CDR strategies more than solving problems — a narrower finding that bears on this claim
• Closed-book self-explanation prompts show no added value on delayed posttest performance, even with prior active processing of instructional explanations — a narrower finding that bears on this claim
• Frequent use of reflective prompts related to better strategic problem-solving and conceptual understanding — a narrower finding that bears on this claim
• Deep understanding is characterized by coherent, interconnected mental schemas that make knowledge transferable — related
• The worked example effect is explained by reduced element interactivity relative to problem solving, and reverses with expertise — related
• Adding a self-explanation feature to a base game improved transfer performance — a narrower finding that bears on this claim
• Generated self-explanations mediate the immediate posttest advantage of open-book over closed-book prompts — a narrower finding that bears on this claim
• The effectiveness of each generative learning strategy is subject to potential boundary conditions discussed in the review — related
• Static self-explanation prompts in the Geometry Cognitive Tutor produced greater gains on Reason and Not Enough Info test items than problem-solving practice alone — a narrower finding that bears on this claim
• High prior knowledge students learned most conceptual understanding from questions plus arguments; answering questions alone was least effective — related
• Self-explanation substantially increases time on task, raising questions about its efficiency — related
• Backward fading and principle prompts supported transfer in probability learning — related
• Prompts and scaffolds for paraphrasing, criticizing, questioning, and synthesizing arguments produce significantly more domain-specific and domain-general knowledge — related
• Experts introduce procedure and planning explanations over time, typically after an initial step or action is completed — related
• Adding instructional explanations to worked examples yields a small positive effect on learning, stronger for conceptual than procedural knowledge — related
• Self-explanation benefits learning across ages, but evidence is mixed for elementary-school children and effects shrink when time on task is equated — related
• Learning from examples depends on the explaining mode: prompting principle-based self-explanations and guided comparison of examples improves transfer — a narrower finding that bears on this claim
• Provided visuals can act as a crutch: explaining with provided visuals yielded higher-quality explanations but no transfer advantage over explaining without them — related