Elaboration and self-explanation are learning processes in which students construct meaning by explaining relationships, causes, principles, and solution steps. Elaboration connects new information to relevant prior knowledge, often through examples, comparisons, or explanatory questions. Self-explanation is a more specific generative activity: the learner produces an account of why a statement is true, why a procedure works, how one idea follows from another, or how new material fits an existing conceptual model. Their educational value lies in improving the quality of the learner's mental representation, particularly when the explanation is accurate, focused, and tested against evidence. Chi et al., 1989; Fiorella, 2023.
The essential practical question is how explaining changes understanding. A student may correctly repeat that a closed electrical circuit allows current to flow, yet struggle to explain why a broken connection interrupts that flow. Explaining requires the student to identify the underlying relation, account for the relevant conditions, and use the principle to predict what will happen in a new situation. Such reasoning can reveal a knowledge gap that recognition or repetition conceals. In experimental work, prompted self-explanation has improved learning on average, while the size and durability of its benefits depend on the materials, learners, supports, comparison activity, and outcome measures. Bisra et al., 2018; Rittle-Johnson et al., 2017.
What Elaboration and Self-Explanation Mean
Elaboration is a broad family of cognitive operations through which a learner enriches new material with meaningful associations. The new content becomes linked to causal knowledge, existing categories, contrasting cases, concrete examples, explanatory principles, or relevant experiences. Useful elaboration is selective: a connection matters because it clarifies the target concept or provides a justified path for remembering and applying it. An imaginative association that misrepresents the concept can be memorable and still be educationally misleading. Dunlosky et al., 2013.
Self-explanation asks the learner to make implicit reasoning explicit. Someone studying a solved equation can explain why a term moves between sides only when the same operation preserves equality. Someone reading a biology text can explain how the components of a circulatory system support its function. The explanation can be spoken, written, drawn with annotations, or expressed through a sequence of justified choices. The critical feature is construction of relevant inferences, not the number of words produced. Chi et al., 1994; Renkl, 1997.
Elaboration, elaborative interrogation, and self-explanation
Elaborative interrogation is a specific prompting method, often built around questions such as “Why does this fact make sense?” or “Why would this principle produce that result?” A well-formulated answer can activate useful prior knowledge and build a causal interpretation. Self-explanation covers a wider set of inferences: explaining a worked step, reconciling apparent contradictions, inferring a missing connection, stating the purpose of a diagram component, or predicting the consequences of altering a condition. The two methods overlap, and their effects should not be presumed identical. In an experiment involving 55 university students learning cardiovascular facts, self-explanation outperformed elaborative interrogation and repetition on the studied recall and recognition outcomes; elaborative interrogation did not outperform repetition in that particular design. This is evidence of a context-dependent comparison, not a universal ranking. O'Reilly et al., 1998.
Summarization describes what material says, whereas a strong self-explanation makes the organizing relations intelligible. An adequate summary of a mathematical example might report the sequence of computations; a self-explanation states what property licenses each operation. Retrieval practice concerns producing information without access to the answer, while self-explanation concerns constructing its meaning. A student can retrieve an answer and then explain it, or explain an example before attempting unaided retrieval. These activities can be combined because they address different dimensions of learning. For the distinct evidence on long-term testing effects, see Retrieval Practice: Why Testing Yourself Improves Memory. Dunlosky et al., 2013.
From Early Cognitive Research to Generative Learning
The modern self-explanation literature emerged from work on how learners study worked-out problems. Chi, Bassok, Lewis, Reimann, and Glaser examined students thinking aloud while studying mechanics examples. Students who produced richer explanations connected steps to underlying principles and displayed knowledge that was less dependent on the original examples. This influential study established an association between explanation quality and understanding and helped formulate the self-explanation effect. Its protocols also made visible the role of monitoring: productive learners noticed when an explanation was incomplete or inconsistent. Because spontaneous explanation quality was observed rather than experimentally assigned, this initial result alone cannot establish the causal size of the technique. Chi et al., 1989.
Subsequent studies directly prompted students to explain. In a small study of eighth graders reading about the circulatory system, 14 students were instructed to explain lines of text and 10 comparison students read the passage twice. The prompted group showed greater gains on the study's knowledge measures, and high-quality explainers demonstrated stronger conceptual models. That experiment provided an important causal test, with the limited sample and immediate assessment defining how far the finding can be generalized. Chi et al., 1994.
The broader generative-learning tradition describes learning as selecting important material, organizing it into a coherent representation, and integrating it with relevant knowledge. A 2023 review by Fiorella organizes sense-making activities around explaining, visualizing, and enacting, arguing that representations and actions can make higher-quality explanations possible. This framework identifies plausible mechanisms and synthesizes experimental evidence, while the precise causal contribution of each proposed mental process remains an empirical question. Fiorella, 2023.
How Explaining Can Build Understanding
Generating inferences and filling conceptual gaps
Instructional materials frequently omit connections an expert assumes are obvious. A physics solution may show a substitution without stating why the selected formula applies. A history chapter may place events in sequence without explicitly deriving the political mechanism connecting them. Self-explanation prompts the learner to produce these missing bridges. Generating a defensible bridge requires locating relevant prior knowledge, checking whether it applies, and revising the interpretation when the pieces fail to fit. Evidence from worked examples and expository text links these processes with improved explanatory understanding. Chi et al., 1989; Chi et al., 1994.
The product of explaining is often a model of relations rather than a longer list of facts. In biology, a learner connects structure, pressure, flow, and oxygen transport. In mathematics, the learner connects a symbolic manipulation with an invariant property. In social science, the learner distinguishes a causal hypothesis from the observation that two outcomes co-occur. A relational model gives the learner something to test against cases that differ from the original illustration. Fiorella, 2023.
Connecting new material with accurate prior knowledge
Elaboration depends on the knowledge brought to the task. A familiar, scientifically sound principle can support a new inference; an inaccurate assumption can produce a coherent-sounding but mistaken explanation. The instructional problem is therefore to activate knowledge that is both relevant and reliable. Worked examples, diagrams, definitions, comparisons, and corrective feedback can supply missing premises. Experiments on prior knowledge and the study of worked examples support the importance of carefully matched guidance. Renkl, 1997; Leppink et al., 2012.
A productive explanation makes its assumptions visible. “The larger number always means the larger fraction” is a plausible but faulty heuristic; comparing one-half with one-third shows why denominator size cannot be interpreted in isolation. The student should identify the size of the equal parts, test the explanation with a visual model, and restate the relationship. In such cases, the value of explaining is inseparable from access to examples or feedback capable of exposing an error. Rittle-Johnson et al., 2017.
Monitoring comprehension and repairing errors
As a learner tries to explain, missing steps become easier to detect: “I know the next line follows, but I cannot justify why.” That observation can trigger rereading, seeking a definition, comparing two cases, asking a teacher, or checking a worked solution. This interaction between explanation and monitoring is compatible with the classic findings on successful self-explainers. Yet an explanation can be articulate and false, so subjective clarity and verbal fluency cannot serve as the only measure of understanding. Chi et al., 1989; Dunlosky et al., 2013.
Abstraction and transfer
Explaining a solution can reveal which features are necessary and which are incidental. For example, a student learning proportional reasoning should recognize that the multiplicative relationship matters across different currencies, quantities, and contexts; the particular numbers are secondary. Several worked-example experiments investigated whether elicited explanations and example variation contribute to transferable understanding. Positive results support a reasonable instructional aim, although transfer effects vary by domain, problem novelty, testing delay, and available scaffolding. Renkl et al., 1998; Rittle-Johnson et al., 2017.
What the Evidence Shows
Broad meta-analysis of self-explanation prompts
Bisra, Liu, Nesbit, Salimi, and Winne synthesized 69 effect sizes from 64 research reports concerning prompts to self-explain during study or problem solving. Their random-effects meta-analysis reported an overall mean effect of Hedges' g = 0.55, indicating a moderate average advantage on the learning measures used in the included studies. The analysis considered differences in learning task, subject matter, education level, inducement method, and intervention duration. This is substantial evidence that prompting explanation can help, while a pooled standardized effect does not establish that every learner, classroom, or long-term outcome will improve by the same amount. Bisra et al., 2018.
A meta-analytic average must be interpreted alongside the designs from which it is assembled. Studies can compare explaining with repeated reading, ordinary study, different activities, or different quantities of guidance. Their outcome measures can include immediate tests, conceptual questions, procedural tasks, and measures at different levels of difficulty. Pooling these results answers whether a strategy tends to help across selected conditions, but it does not identify a single mechanism, guarantee an independent effect on delayed retention, or predict the exact payoff of a particular teaching plan. The comparator and outcome are essential parts of any claim about effectiveness. Bisra et al., 2018; Dunlosky et al., 2013.
Mathematics-specific synthesis
Rittle-Johnson, Loehr, and Durkin reviewed the mathematics evidence and found small-to-moderate positive effects of prompted self-explanation on immediately measured procedural and conceptual knowledge and procedural transfer. The review emphasized instructional design and reported more limited evidence for reliable benefits in authentic classrooms or for retaining knowledge after a delay. This distinction prevents treating a good immediate problem-solving score as proof of durable mastery. Rittle-Johnson et al., 2017.
Comparative judgments about study techniques
In their broad review of ten study techniques, Dunlosky and colleagues assigned elaborative interrogation and self-explanation a moderate-utility assessment. Their judgment reflected promising benefits alongside limitations in breadth and educational-context evidence. Practice testing and distributed practice received stronger general ratings across the criteria assessed. These are assessments of the research literature available to the review, rather than a claim that every form of explaining is inferior to every form of testing. Studying to understand a causal mechanism and studying to reproduce it after a week pose related but different educational demands. Dunlosky et al., 2013.
Evidence established, preliminary, and limited
An established conclusion is that prompting students to construct relevant explanations can improve learning relative to particular comparison conditions, with benefits supported by experiments and quantitative syntheses. A supported but more conditional conclusion is that explanations enhance conceptual transfer when learners connect examples to applicable principles and receive appropriate guidance. Evidence is more limited for the best universal prompt, exact dosage, durable delayed effects across every field, and routine large-scale classroom implementation. Competing accounts of why the intervention works include deeper inference generation, increased time and attention, improved error detection, and interaction with supplied examples; measured outcomes do not always isolate these components. Bisra et al., 2018; Fiorella, 2023; Rittle-Johnson et al., 2017.
Different Learning Domains
Mathematics and worked examples
Mathematics is a natural setting for step-by-step justification because operations and representations have explicit rules. A beginner may see a completed equation and imitate the surface pattern; explanation asks which equality-preserving operation is being performed and why. Renkl's work found important individual differences in how learners study probability examples. Further experimental research examined example variation and induced self-explanations as supports for transfer. These findings justify asking students to connect a worked line with the governing principle, while providing correct examples and feedback when knowledge is incomplete. Renkl, 1997; Renkl et al., 1998.
For example, suppose a worked solution transforms 3(x + 2) = 18 into x + 2 = 6. A superficial description is “divide by three.” A self-explanation identifies that both sides are divided by the same nonzero quantity, preserving equality, and explains why the subsequent subtraction is valid. The student then checks the reasoning by substituting x = 4 into the original equation. A stronger transfer question changes the coefficients and asks what principle remains constant. This teaching example illustrates the mechanism; it is not a result from a particular experiment.
Science and causal systems
Scientific explanation often requires identifying interactions among parts of a system. The classic circulatory-system study used prompted explanations to help students connect textual statements to a more coherent model of function. A later experiment comparing self-explanation, elaborative interrogation, and repetition illustrated that the wording and cognitive demands of the prompt can affect recall and recognition. These findings support questions that ask learners to identify mechanisms, explain anomalies, and predict changes instead of repeating isolated facts. Chi et al., 1994; O'Reilly et al., 1998.
Consider the statement “Gas exchange occurs across thin alveolar walls.” A weak response repeats that thin walls help oxygen move. A stronger explanation connects diffusion distance with the movement of molecules down a concentration gradient, specifies the relevant structure, and predicts why thicker walls could impair exchange. The explanation should be checked against a scientifically accurate representation, because physiological outcomes depend on multiple interacting conditions. The distinction between a causal chain and a memorized phrase is the educational target.
Reading and comprehension
When reading expository text, learners may understand individual sentences while failing to connect them into a coherent account. Self-explanation can prompt bridging inferences, identification of referents, integration with background knowledge, and detection of contradictions. McNamara's Self-Explanation Reading Training examined the effects of teaching strategies for better explanation while reading science passages. The results included variation with text difficulty and reader knowledge, showing the relevance of instruction rather than merely telling students to “explain more.” Later work also investigated transfer from reading-strategy training to science-course performance. McNamara, 2004; McNamara, 2017.
For a dense paragraph, a usable sequence is to identify the main proposition, state how it follows from what came before, explain unfamiliar terms, and test the account with a concrete example. If the paragraph reports a correlational research finding, the learner should explain why association alone does not identify the cause. If the text contains a conditional claim, the learner should state the conditions under which it holds. These responses reveal comprehension more clearly than copying a sentence with synonyms.
History, social science, and conceptual subjects
In historical or social-science study, elaboration can organize evidence around mechanisms and alternative explanations. A student explaining a policy change might describe material incentives, institutional constraints, and actors' beliefs, then ask which observations would distinguish these accounts. This use is theoretically compatible with generative learning, but effects observed in math or science laboratories cannot be automatically assigned to history classrooms. Evaluation requires domain-appropriate comparisons and assessments that reward evidentiary reasoning rather than length of response. Fiorella, 2023; Dunlosky et al., 2013.
When Explanation Helps Less
Inadequate prior knowledge and cognitive load
To generate a valid explanation, learners must have enough background information to identify the relevant principle. When a task is already difficult, asking a novice to justify every step can exhaust attention while producing guesses. Research on expertise reversal in statistics found that instructional approaches differ in their effects for learners with different levels of prior knowledge. A reasonable teaching design therefore supplies more complete explanations and examples early, then transfers responsibility to learners as competence grows. Leppink et al., 2012; Renkl, 2002.
The same prompt is not equally informative at every stage. An experienced student can explain why a mathematical shortcut works; a novice may first need a labeled demonstration and a clear statement of the governing theorem. Guidance can then fade from “Here is the principle; explain where it is applied” to “Explain your choice and justify each transformation.” This sequence represents an instructional design principle, not an inflexible age-based prescription. Renkl et al., 1998; Fiorella, 2023.
False explanations and seductive coherence
People can produce explanations that are fluent, elaborate, and incorrect. In classrooms, explanation quality should be evaluated against concepts, worked solutions, trusted sources, or empirical observations. A confident statement that a heavier object always falls faster, for instance, needs a clearly specified physical model and experimental evidence before it is accepted. Self-explanation can reveal the mistaken assumption, but it cannot repair the assumption without corrective information. The importance of feedback is also reflected in research on coupling explanation with worked examples. Renkl, 2002; Fiorella, 2023.
Excessive prompting and fragmented study
Prompts placed after every sentence or operation can interrupt the organization of a larger argument. If students respond mechanically, the task becomes compliance with a question format rather than construction of a coherent mental model. Good design selects conceptually important transitions, contrasts, anomalies, and solution decisions. The appropriate frequency depends on complexity, learner expertise, and whether the immediate goal is initial comprehension, error correction, or later practice. Fiorella, 2023; Bisra et al., 2018.
Designing Better Self-Explanation Prompts
Useful prompts ask for a relationship that can be evaluated. “Why does the method work here?” directs attention to the principle behind an operation. “Which assumption allows this inference?” asks for a premise. “What would change if the condition were removed?” tests the mental model. “How does this example differ from the previous case?” encourages discrimination. “Where does the explanation conflict with the data?” invites revision. These are design examples informed by evidence on generative activity and worked-example learning, not a claim that every wording has been separately validated in a randomized trial. Fiorella, 2023; Renkl et al., 1998.
A complete prompt sequence can begin with a correct model, ask the learner to explain one consequential step, compare the explanation with a criterion, and then present a new problem. For beginning learners, the criterion may be an annotated worked example. For advanced learners, it may be a theorem, a research result, or a peer critique grounded in evidence. When explanation is assigned as homework, instructors should indicate what constitutes an acceptable explanation: naming the principle, identifying the mechanism, accounting for the relevant conditions, and checking the prediction. Renkl, 2002; Rittle-Johnson et al., 2017.
A worked example that elicits reasoning
An instructor teaching proportional reasoning might show that a recipe requiring 250 grams of flour for five portions requires 400 grams for eight portions. Before asking students to repeat the arithmetic, the instructor asks, “Which quantity remains constant when the number of portions changes?” A sound explanation states that flour per portion stays constant, so multiplying 250 by 8/5 preserves the ratio. The follow-up asks whether the same ratio applies if the recipe's composition changes; it does not, because the initial proportional condition has been altered. This sequence moves from a procedural line to a conceptual condition and a transfer check.
The instructor should compare responses against the mathematical structure. “I multiplied because the second number was bigger” is insufficient; it might lead to the right answer in this instance without showing an understanding of proportionality. Clear evaluation prevents explanation length from being mistaken for explanation quality. This application draws on the worked-example literature rather than constituting a new empirical intervention. Renkl, 1997; Rittle-Johnson et al., 2017.
How Students Can Use Elaboration While Studying
Begin by reading or watching an accurate explanation of a manageable unit, then identify its central claim. State the claim in your own words and explain why it follows, explicitly naming the mechanism, condition, or governing rule. Add a new example that differs meaningfully from the original. Compare your explanation with the instructional material or an authoritative answer, correct an error, and attempt a fresh question without the original explanation visible. The final unaided step helps establish whether understanding is available independently of the original prompt. This is a synthesis of evidence-backed principles, rather than a protocol whose exact steps and timing have all been directly tested together. Dunlosky et al., 2013; Fiorella, 2023.
For a chapter on ecosystems, a student can explain why removing a predator may alter more than one population, draw a food-web relation, and predict how a second environmental change might modify the result. For a statistics example, a student can explain why a confidence interval is tied to a method and sampling conditions rather than calling it a range that “contains all likely values.” For a language concept, a student can compare two grammatical constructions and identify the contextual condition governing their use. Each activity should end with a factual check, because the goal is an accurate model that supports future decisions.
Elaboration can fit alongside a longer study schedule. After initial comprehension, attempting to reconstruct the principle without looking at the notes provides a retrieval opportunity; returning to it after a delay tests retention. The independent evidence on temporal scheduling belongs to Spaced Repetition: How the Spacing Effect Supports Long-Term Memory. Explanation adds conceptual structure, while a later attempt tests whether the structure remains accessible. Dunlosky et al., 2013.
Classroom Instruction, Feedback, and Peer Explanations
Teachers can model a high-quality explanation before asking students to supply one. A model makes clear the difference between repeating a step and justifying it. Students can then annotate a worked example, explain a deliberate error, compare alternative solutions, or discuss why an unexpected observation follows from a principle. Written explanations can be short if they identify the decisive causal or logical connection. The evidence for self-explanation offers support for these structured activities, while effects on grades and learning across a full school year still require context-specific evaluation. Bisra et al., 2018; Renkl, 2002.
Peer discussion adds a distinct dimension: a learner must make reasoning intelligible to another person and respond to questions. These social demands can reveal assumptions and stimulate revision, yet benefits from peer explanation depend on interaction quality, content accuracy, and the activity's structure. Findings about individually generated self-explanations cannot by themselves establish a causal advantage for peer teaching. A review of learning by teaching describes both its promise and the importance of what learners actually do while preparing, explaining, and responding to others. Duran, 2017.
Feedback should evaluate the reasoning. Instead of responding only “correct” or “incorrect,” an educator can identify the step at which a premise fails, contrast the student's explanation with the applicable rule, and ask for a revised prediction. This supports the sequence of noticing, correcting, and applying an explanation. Corrective feedback is particularly important where students are initially learning concepts whose surface plausibility makes misconceptions resistant to ordinary rereading. Renkl, 2002.
Learner Differences and Educational Access
A learner's prior knowledge, reading fluency, language proficiency, working-memory demands, and familiarity with disciplinary conventions can affect the effort required to articulate an explanation. The same task can ask one student to connect familiar ideas and another to construct vocabulary and concepts simultaneously. Teachers should therefore allow multiple response modes when those modes can demonstrate the intended reasoning, provide language scaffolds where appropriate, and separate assessment of conceptual understanding from irrelevant demands of verbal performance. This is an instructional inference grounded in evidence for learner variability and expertise-sensitive supports, not a claim that a single accommodation has a universal measured effect. Leppink et al., 2012; Fiorella, 2023.
Students with less background knowledge may benefit from partially completed explanations, vocabulary aids, labeled diagrams, and opportunities to ask specific questions. More advanced learners may benefit from fewer prompts and more open-ended transfer problems. A student can convey valid reasoning orally, in a paragraph, through an annotated diagram, or by linking mathematical representations. Educational accessibility depends on preserving the intellectual demand while reducing barriers that are extraneous to the skill being assessed. These implementation choices should be assessed for the actual learner group rather than inferred from a diagnostic label alone.
Explanation, Digital Tools, and Generative AI
Digital platforms can deliver prompts, compare solutions with criteria, and collect learner reasoning. Their educational contribution depends on whether students still perform the inferential work and whether feedback is accurate. A system that immediately supplies a polished explanation may help a novice inspect a correct model, but it can also remove the demand to generate the critical inference. The distinction between observing an answer and independently reconstructing its logic is central to generative learning research. Fiorella, 2023.
With a generative AI tool, a learner can request a worked example and then close the answer, explain its steps without looking, identify assumptions, and solve an analogous problem. The system can produce contrasting cases or candidate questions, while the student checks scientific statements and calculations against trustworthy references. Because generated explanations can be wrong, evaluation of a particular AI tutoring implementation requires independent measures of accuracy, learner agency, delayed retention, transfer, and educational privacy. The existing meta-analysis of human-designed self-explanation prompts does not establish the efficacy of commercial AI tutors. Bisra et al., 2018; Fiorella, 2023.
A useful boundary is to keep explanation production with the learner. Software can ask “Which principle justifies step three?” and help locate the relevant source; the learner should state the principle, test it, and revise it. In educational research, time saved, satisfaction, or assignment completion should be distinguished from improved independently measured learning. Digital access and protection of student responses also matter when platforms store work, behavioral traces, or sensitive educational data. These considerations are design and evaluation requirements, not empirical proof of one technology's effectiveness.
Measuring Understanding and Research Quality
Explanations can be scored for factual correctness, relevance, causal or logical coherence, coverage of necessary conditions, and responsiveness to contradictory evidence. A scoring rubric should match the subject. In mathematics it may ask whether a principle licenses each transformation; in science it may require a causal chain; in history it may require explicit evidence and treatment of alternatives. Length, sophisticated vocabulary, and verbal confidence are poor substitutes for these dimensions. Think-aloud protocols reveal processes during learning but can also change the task, and their interpretation requires careful coding. Chi et al., 1989; Renkl, 1997.
A rigorous intervention study should distinguish immediate comprehension, delayed retention, and transfer to new problems. It should specify what comparison students do, how much study time each group receives, whether assessors are blinded when feasible, and whether outcomes were selected before results were known. Random assignment strengthens causal inference, while repeated classroom implementation helps assess ecological validity. Larger preregistered replications, transparent reporting of null findings, and analysis of effects by prior knowledge would clarify how to use prompts under real educational constraints. The existing syntheses show why average effects need to be interpreted against heterogeneity and study designs. Bisra et al., 2018; Rittle-Johnson et al., 2017.
A powerful evaluation asks whether the learner can explain a genuinely new case after a delay without receiving the original worked solution. This measure tests for generalizable reasoning more directly than asking whether the initial paragraph sounded clear or whether a familiar exercise was completed quickly. Classroom assessments can combine correct-answer scoring with short justifications, comparisons, and prediction questions, provided the scoring criteria are reliable and accessible to the intended learners.
Frequently Asked Questions
Is self-explanation the same as thinking aloud?
Thinking aloud is a way of reporting thoughts during a task and can be used as a research method. Self-explanation is a particular kind of constructive reasoning: the learner attempts to explain a relation, principle, or step. A person can think aloud while merely describing what is visible; a student can also self-explain silently or in writing. The relevant learning activity is the quality of the inference, not the act of speaking. Chi et al., 1989.
Is asking “why?” enough?
A why-question can be useful when it elicits an accurate causal or conceptual account. It can be unproductive if the learner lacks the necessary knowledge or supplies a guess that is never evaluated. Questions about assumptions, mechanisms, comparisons, and predictions may expose different aspects of understanding. Experimentally, methods grouped under elaborative interrogation and self-explanation have not always produced identical outcomes. O'Reilly et al., 1998.
Does explaining guarantee better long-term memory?
The available evidence supports learning benefits under many conditions, but long-term retention findings are less comprehensive than immediate learning findings in important subdomains. To establish durable knowledge, assess performance after a meaningful delay and consider complementary retrieval and spacing practices. A successful immediate explanation is a valuable sign of current understanding, not a guarantee that the explanation will remain available weeks later. Rittle-Johnson et al., 2017; Dunlosky et al., 2013.
Should students explain every step?
Students benefit from explaining decisive principles, transitions, and errors. Novices can need explicit models, while knowledgeable learners may profit from less intrusive prompts. Continuous explanation of trivial or obvious actions may impose demands without adding conceptual value. Prompt selection should reflect prior knowledge and the complexity of the material. Leppink et al., 2012; Fiorella, 2023.
Can teaching someone else replace self-explanation?
Teaching can create opportunities to organize and explain knowledge, and questions from another person can reveal gaps. The effects depend on preparation, accuracy, the listener's contribution, and how the learning outcome is assessed. Self-explanation, peer explanation, and learning by teaching overlap in cognitive demands but constitute different interventions that need their own evidence. Duran, 2017.
Scientific Conclusion
Elaboration and self-explanation build understanding when they turn new information into accurate relationships that the learner can justify, test, and apply. The scientific evidence is strongest for prompted explanations improving measured learning under defined conditions, including worked examples, expository texts, and mathematical reasoning. The most consequential qualifications concern explanatory accuracy, prior knowledge, task demands, assessment delay, and transfer. Instruction works best when it supplies the knowledge needed for valid inference, invites learners to articulate consequential steps, checks their reasoning against evidence, and evaluates understanding independently of the original example. Bisra et al., 2018; Rittle-Johnson et al., 2017; Fiorella, 2023.
Related Articles
Cluster Learning and Educational Psychology — The scientific parent chapter situating generative learning among educational psychology theories, methods, and practices.
Retrieval Practice: Why Testing Yourself Improves Memory — The independent account of remembering through testing and its evidence for retention.
Spaced Repetition: How the Spacing Effect Supports Long-Term Memory — The separate account of distributed study and long-term retention.
Interleaving Practice: When Mixing Subjects Improves Learning — Compare interleaving with blocked practice, including discrimination learning and when effects vary.
Note-Taking Methods: Evidence on Learning and Retention — Review written and digital notes, generative processing, note review, retention, and context differences.
How to Study Effectively: Evidence-Based Learning Strategies — Compare effective learning strategies, practice schedules, elaboration, and planning while referencing existing memory canonicals.
Deliberate Practice in Skill Learning: Expertise, Feedback, and Limits — Examine purposeful repetition, corrective feedback, expertise, and contested estimates of practice effects.
References
Bisra, K., Liu, Q., Nesbit, J. C., Salimi, F., and Winne, P. H. (2018). Inducing self-explanation: A meta-analysis. Educational Psychology Review, 30, 703–725. https://doi.org/10.1007/s10648-018-9434-x
Chi, M. T. H., Bassok, M., Lewis, M. W., Reimann, P., and Glaser, R. (1989). Self-explanations: How students study and use examples in learning to solve problems. Cognitive Science, 13, 145–182. https://doi.org/10.1207/s15516709cog1302_1
Chi, M. T. H., De Leeuw, N., Chiu, M.-H., and Lavancher, C. (1994). Eliciting self-explanations improves understanding. Cognitive Science, 18, 439–477. https://doi.org/10.1207/s15516709cog1803_3
Dunlosky, J., Rawson, K. A., Marsh, E. J., Nathan, M. J., and Willingham, D. T. (2013). Improving students' learning with effective learning techniques: Promising directions from cognitive and educational psychology. Psychological Science in the Public Interest, 14(1), 4–58. https://doi.org/10.1177/1529100612453266
Duran, D. (2017). Learning-by-teaching: Evidence and implications as a pedagogical mechanism. Innovations in Education and Teaching International, 54(5), 476–484. https://doi.org/10.1080/14703297.2016.1156011
Fiorella, L. (2023). Making sense of generative learning. Educational Psychology Review, 35, Article 50. https://doi.org/10.1007/s10648-023-09769-7
Leppink, J., Broers, N. J., Imbos, T., van der Vleuten, C. P. M., and Berger, M. P. F. (2012). Self-explanation in the domain of statistics: An expertise reversal effect. Higher Education, 63, 771–785. https://doi.org/10.1007/s10734-011-9476-1
McNamara, D. S. (2004). SERT: Self-explanation reading training. Discourse Processes, 38(1), 1–30. https://doi.org/10.1207/s15326950dp3801_1
McNamara, D. S. (2017). Self-explanation and reading strategy training (SERT) improves low-knowledge students' science course performance. Discourse Processes, 54(7), 479–492. https://doi.org/10.1080/0163853X.2015.1101328
O'Reilly, T., Symons, S., and MacLatchy-Gaudet, H. (1998). A comparison of self-explanation and elaborative interrogation. Contemporary Educational Psychology, 23(4), 434–445. https://doi.org/10.1006/ceps.1997.0977
Renkl, A. (1997). Learning from worked-out examples: A study on individual differences. Cognitive Science, 21(1), 1–29. https://doi.org/10.1207/s15516709cog2101_1
Renkl, A. (2002). Worked-out examples: Instructional explanations support learning by self-explanations. Learning and Instruction, 12(5), 529–556. https://doi.org/10.1016/S0959-4752(01)00030-5
Renkl, A., Stark, R., Gruber, H., and Mandl, H. (1998). Learning from worked-out examples: The effects of example variability and elicited self-explanations. Contemporary Educational Psychology, 23(1), 90–108. https://doi.org/10.1006/ceps.1997.0959
Rittle-Johnson, B., Loehr, A. M., and Durkin, K. (2017). Promoting self-explanation to improve mathematics learning: A meta-analysis and instructional design principles. ZDM Mathematics Education, 49, 599–611. https://doi.org/10.1007/s11858-017-0834-z