Skip to content

Develop Understanding

Pattern · All patterns
Evidence · 28 claims (13 for, 13 mixed, 2 against) · 42 studies (18 causal, 9 review, 8 quant-synthesis, 4 theoretical, 2 qualitative, 1 associational), q1–q4 · 8 of 42 report an effect size · 20 claims rest on one study

Description and scope

A reusable lesson policy for teaching a concept or principle so that learners can explain it and use it on new cases (conceptual-understanding, near-transfer), not only recall it: elicit what learners already think about the idea → explain and represent the new idea, naming any common wrong idea and saying why it fails → guide learners through examples and short applications while they explain each step to themselves → check understanding on a changed item and repair what the check shows → connect the idea to contrasting cases, related ideas and later use. Over several sessions, support moves from teacher-supplied examples to learner-generated ones, and the check moves from the next item to a changed item a week or more later.

It sits beside two converted patterns and does not repeat their models. Gagné's nine events is the general lesson architecture for any structured target; this page is a five-step specialisation of it for a conceptual target, where the distinctive work is surfacing and confronting prior ideas and getting learners to explain and connect. The Direct Instruction pattern is the sibling policy for teaching a structured target explicitly and handing over on unaided success; this pattern uses the same explain–guide–check spine and adds explanation prompts, misconception repair and connection-making. Single steps are governed by converted principles: Activation for step 1 and Guided Practice for step 3. For a target that must be induced from examples and non-examples, use Concept Attainment; for a complex-skill practised as a whole, see Merrill's First Principles of Instruction.

No claim in this wiki tests Develop Understanding as a sequence, against a lesson that explains and then sets independent practice, or against anything else. The claims that bear on it test single components: prompting learners to self-explain against not prompting them; refutation or conflict-producing instruction against standard exposition for learners holding a misconception; explicit or guided instruction against unassisted discovery; comparing cases side by side against studying them one at a time; contingent against fixed support. Each is carried to this pattern by extrapolation, and the page says how far. The sequence below is the earlier page's, kept as the default, with each step's evidence status stated; the response-dependent branches are an untested design proposal.

Inputs: establish the design brief

Field Record or elicit
Learner and starting observation Before instruction, a prediction or explanation of one familiar case the idea governs ("what will happen, and why?"), with the reasoning given and how confidently it was offered. Record task-specific expertise as observed (novice, intermediate, advanced, or ?), age band (child, adolescent, adult) and setting. A blank answer does not show that nothing is known; a fluent one does not show that it is accurate.
The idea and its known wrong versions What the concept or principle is, the relationship or reason a learner should be able to state, and the common wrong ideas about it (from earlier classes, the subject's literature, or the starting responses). Whether element interactivity is low or high for these learners.
Context and activities Setting (classroom, online-self-paced, online-instructor-led, workplace-clinical), single-session or days-weeks, lesson length, group size, whether someone can read learners' explanations during the lesson and respond, and which representations are available (diagram, simulation, worked case, text).
Learner-valued goal Ask what the learner wants to be able to do with the idea (explain it to someone, use it at work, pass an exam, make sense of something they noticed). Record disagreement with the designer objective.
Designer objective Which capability the assessment will weigh: stating the relationship, explaining why a case comes out as it does, predicting a new case, choosing between this idea and a neighbouring one, or using it in a later task.
Outcome An item the learner has not seen, in a changed representation or context, asking for a prediction and a reason, scored apart from recall of the definition; immediate or delayed; and whether the idea must be used inside a richer task (near-transfer, far-transfer).

A request to "help learners understand X" does not specify these. Ask for the idea, the wrong ideas likely to be held, the starting response and the test item before choosing the examples, the prompts or the number of sessions.

Sequence and conditional policy

The default sequence keeps the earlier page's five steps; the label after each says what supports it. Timings are for one 50-minute lesson and are untested proposals.

  1. Elicit what learners already think (5–8 minutes). Ask for a prediction or explanation of a familiar case the new idea governs, in writing and privately before any discussion, so wrong ideas surface rather than hide behind the confident few. Ask for the relation the lesson will use, not for "everything you know about the topic". Evidence status: activation of topic knowledge on its own has often shown little effect (advance organizers and topic activation [~M]: an organizer meta-analysis was favourable, and a randomized experiment with 88 primary pupils found no comprehension gain from activating topic knowledge, while activating metacognitive knowledge helped). Its purpose here is to find the wrong ideas step 2 must confront. Governed by the Activation principle.
  2. Explain and represent the idea, and refute the wrong one (10–15 minutes). Give a clear explanation with one representation that shows the relationship (a diagram, a worked case, a simulation run). Where step 1 or the subject's literature shows a common wrong idea, name it, show a case where it predicts the wrong result, and explain why the correct idea predicts the right one. Evidence status: standard exposition rarely dislodges a misconception, while refutation text and conflict-producing instruction more often do [+M] (a meta-analysis and a review in science and reading education, read from abstracts, no pooled effect recorded). For novice learners, explaining the idea rather than leaving them to find it is supported by explicit instruction against unassisted discovery [+S].
  3. Guide learners through examples and short applications while they explain each step (15–20 minutes). Two or three worked cases, then two or three cases for learners to work, each with a prompt to explain why the step or outcome follows ("which part of the idea tells you that?"). Prompts name the principle or the gap, not "explain this to yourself". Help is given in response to what the explanation shows: a cue, then a more specific hint, then a demonstration. Evidence status: prompting learners to self-explain while studying or solving [+S] (a meta-analysis of 69 effect sizes, g = .55, across task types, subjects and levels; evidence for classroom settings and delayed tests is much more limited, per a mathematics meta-analysis). Contingent support [+M] beat fixed or no support in one-to-one tutoring of children; untested at class scale. How the guided round is run is governed by the Guided Practice principle.
  4. Check understanding on a changed item and repair (5–10 minutes). One item in a representation or context not used in steps 2–3, asking for a prediction and a reason, answered by every learner before any discussion. Repair what the answers show (rows below), and re-check with a second changed item. Evidence status: untested as a step here; from the earlier page. The practice-present experiment [+M] supports keeping a performance step in a lesson; it did not test a check of understanding.
  5. Connect the idea (5–10 minutes). Put two cases side by side, one where the idea applies and one where a neighbouring idea applies, or two surface-different cases governed by the same idea, and ask learners what they share and how they differ; then link the idea to an earlier topic and to where it will be used next. Evidence status: comparing cases side by side [+M] (a meta-analysis of 57 experiments, d = .50, larger when learners look for similarities and when the principle comes after the comparison; one experiment with seventh-graders found comparison raised procedural knowledge and flexibility but not conceptual knowledge).

Dose and progression across sessions (untested proposals). Most conceptual targets need more than one lesson. A default arc for a unit of three to five sessions:

  • Session 1: the full five steps, teacher-supplied examples in one representation, explanation prompts with sentence starters ("this happens because…"), one changed-item check at the end.
  • Sessions 2–3: open with two retrieval items on the idea and one prediction item on a new case (5 minutes); then a second representation (if the first was a diagram, a data table or a narrative case); learners explain without sentence starters; add contrasting cases and the cases where the wrong idea most often returns.
  • Final session: learners generate their own example and non-example and explain them to a peer; a changed-context task that uses the idea inside a larger problem.
  • Delayed check: one or two weeks after the last session, an unseen prediction-and-reason item, because a misconception that seems repaired on the day can return on a delayed or unfamiliar item (the misconceptions claim's discussion, not a tested finding).
  • Moving on: a learner moves from guided to independent cases once they give a correct prediction with a correct reason on two consecutive changed items without prompts; one who already does this at step 1 skips steps 2–3 and goes to step 5 or to independent application (expertise reversal [~M]). The threshold is an untested proposal; record it and adjust it locally.

Between steps, choose the next activity provisionally from what the learner does:

Observation Discriminating probe Proposed action, with its uncertainty
A confident, wrong prediction at step 1 Ask for the reasoning, then a prediction for a case where that reasoning gives a clearly wrong answer. If the same idea recurs on a differently worded case, treat it as a held misconception: in step 2, name it and show the discrepant case before giving the correct account. If it does not recur, it may have been a guess; teach as planned. Neither branch is tested.
Correct definition stated, wrong prediction on the changed item Give a near case in the lesson's own representation and one in a new representation. Right in the old representation, wrong in the new: the idea may be tied to one representation; add a second representation and a comparison of the two. Wrong in both: the explanation has not been connected to cases; return to worked cases with explanation prompts. Proposed.
Explanations that restate the step without a reason ("because you add them") Ask which part of the idea licenses the step, or offer two candidate reasons to choose between. If they can choose but not produce a reason, give principle-naming prompts or partial explanations to complete, and fade them later. If they cannot choose either, the idea itself has not been grasped; re-explain with a contrasting case. The self-explanation evidence notes open prompts can produce restatement; the branch is untested.
Correct answers during guided cases, wrong on the unaided changed item A matched item with prompts kept, then one with prompts removed. Fade the prompts in steps and add cases before re-checking; do not repeat the whole explanation by default. Proposed.
Correct prediction and reason at step 1 One harder changed item and a request to say where the idea stops applying. Skip to connection and to cases at the boundary of the idea; guided worked cases may be redundant for this learner (expertise reversal [~M]). One success does not establish the threshold.
Correct on the day, wrong idea back a week later Re-ask the step 1 prediction in a new context; record what happened in between. Add the discrepant case and a spaced retrieval item to the next sessions rather than lengthening the first lesson. Whether that prevents the return is untested here.

These rows are proposals for local observation, untested with learners. A probe is itself a learning exposure, and one contrast does not identify a cause.

Fitting the design to a situation

The three facts that most change the decision:

  • What learners already think about the idea, and whether it is wrong. A coherent wrong idea is not removed by a clear explanation alone, and a learner who already explains the idea needs little of steps 2–3. If unstated, ask: what do learners predict about one familiar case, and why?
  • The learner's task-specific expertise. For a novice, explaining the idea and guiding its use beats leaving them to discover it; for an advanced learner, the same worked cases and prompts can become redundant. If unstated, ask: can they explain one case of this idea now, unaided?
  • Whether anyone can read learners' explanations during the lesson. The explanation prompts and contingent help depend on someone or something responding to what learners write; most contingent-support evidence is one-to-one. If unstated, ask: during guided cases, who reads a learner's explanation before the next case, and how soon?
If the brief says… Then change… Basis
Learners are novice on this idea Keep steps 2–3 in full: explain first, two or three worked cases with explanation prompts and sentence starters before any case learners work alone; give hints in response to their explanations. Explicit instruction against unassisted discovery [+S]; the meta-analysis does not break results down by prior knowledge
Learners are advanced, or explain the idea correctly at step 1 Skip the worked cases; go to boundary cases, comparison with a neighbouring idea, and use in a richer task, with selective feedback. Expertise reversal [~M] (a review of cognitive-load studies; gives no crossover point)
Learners hold a known, confident misconception (common in science: force and motion, seasons, heat) Make step 2 a refutation: state the wrong idea, show a case where it fails, explain why the correct idea succeeds; return to the discrepant case in later sessions and check at a delay. Misconceptions and refutation [+M]; science and reading education, read from abstracts
Learners are child (under 12) Keep explanation prompts short and oral or pointing-based; use concrete, visible cases; check each child's prediction individually rather than by show of hands. Untested proposal; the self-explanation meta-analysis spans levels of education but the claim page does not report results by age
The goal is a procedure the learner must carry out fluently, not an idea to explain Use the Direct Instruction pattern or Guided Practice instead; keep one explanation prompt per worked case at most. Untested; from the earlier page's scope ("not sufficient for highly complex skills alone")
The goal is a complex-skill performed as a whole Embed the idea in a whole task from the start rather than teaching it as a separate lesson first. Whole-task practice [~M] (a design argument, not a study)
The outcome is transfer to new contexts Use at least two representations and two surface-different cases; in step 5, ask learners to say what the cases share before the principle is restated. Comparing cases [+M]; larger effects when learners look for similarities and the principle comes after, smaller on delayed tests
online-self-paced, no teacher Make every prompt require a typed explanation before the next screen; give worked explanations to compare against; include automatically checked prediction items with feedback; end with an unseen changed item. Practice present in a computer-based lesson [+M] (undergraduates, one lesson, horizon not reported); the explanation-comparison step is untested
classroom with one teacher and a large group Collect step 1 predictions and step 4 checks from every learner at once (cards, mini-whiteboards, a poll); address the commonest wrong idea with the whole class, then go to those still wrong. Untested proposal; contingent support [~M] is shown one-to-one, and its classroom-scale effect is not
single-session only Run the five steps once, keep step 5 to one comparison, and record that understanding at a delay was not checked. Untested proposal
High stakes, or the test is weeks away Judge readiness on an unseen changed item answered unaided, not on explanations given with prompts; add the delayed check. Untested proposal; the self-explanation claim notes that evidence for delayed gains is much more limited
Instruction is in the learners' second language, or literacy is low Let learners predict and explain orally, by drawing or by choosing between given reasons before writing them; keep the representation visual; gloss terms the objective does not include. Untested; from the earlier page's guidance to vary representation and complexity

Six of the twelve rows rest on a claim (novice and advanced learners, misconceptions, complex skills, transfer, and online practice), and each carries its claim beyond what was tested, as its Basis cell says: the whole-task row cites a design argument, and the online row a study of practice, not of explanation. The classroom row cites a claim only for its limit; the other five are untested proposals or the earlier page's guidance.

Choosing configurations from evidence

  • Explanation prompts during guided use. Self-explanation improves conceptual understanding and problem-solving performance [+S]: a random-effects meta-analysis of 69 effect sizes from 64 reports found prompting learners to self-explain while studying or solving improved learning outcomes, g = .55, across task types, subject areas and levels of education (read from the abstract; moderators not reported). A mathematics meta-analysis found small-to-moderate immediate gains in procedural knowledge, conceptual knowledge and transfer, with much more limited evidence in classrooms or after a delay. A small experiment with 24 eighth-graders found those prompted to self-explain each line of a circulatory-system text gained more than those who read it twice; an observational study of 10 students found stronger students generated more self-explanations. Not settled against its sources. It supports step 3; it does not test the five-step sequence, and its discussion notes that learners lacking background can explain their way into a wrong idea, which is why step 2 comes first.
  • Confronting a held wrong idea. Instruction that confronts science misconceptions (conceptual-conflict interventions, refutation text) produces conceptual change more often than standard exposition [+M]: a meta-analysis of conceptual-change interventions in science and reading education found that interventions producing conceptual conflict were more effective than standard exposition, and a review of two decades of refutation-text research found refutation text more likely than expository text to produce conceptual change, with no age pattern. Both read from abstracts, no pooled effect recorded. They test interventions that overcome misconceptions, not this lesson; carry them to "where a wrong idea is held, name and refute it", no further.
  • Explaining rather than leaving novices to discover. Minimal guidance is less effective for novices than explicit instruction [+S]: across 580 comparisons explicit instruction outperformed unassisted discovery (d = 0.38), and discovery enhanced with feedback, worked examples, scaffolding or elicited explanations outperformed other instruction (d = 0.30); 112 third- and fourth-graders learned the control-of-variables procedure far more often from direct instruction and transferred it as well. This places the pattern on the guided side: step 1 elicits, it does not ask learners to find the idea.
  • Comparison to connect. Comparing cases side by side improves learning and transfer by a moderate average amount, though in one algebra study it raised procedural knowledge and flexibility but not conceptual knowledge [+M]: a meta-analysis of 57 experiments (336 tests) found case comparison produced more learning than single or sequential case study and other conditions, d = .50, 95% CI [.44, .56], larger when learners looked for similarities, when the principle came after the comparison, for perceptual content and on immediate tests; 128 undergraduates comparing two negotiation cases used the principle later more often than those studying them separately (48% vs 19%); 70 seventh-graders comparing algebra methods gained procedural knowledge and flexibility but not more conceptual knowledge. Supports step 5's side-by-side comparison; qualifies it for a conceptual outcome.
  • Responding to what learners show. Contingent scaffolding improves learning more than fixed or absent support [+M]: fourth and fifth graders tutored one-to-one in long division with fully contingent support outperformed fixed, moderate or no support at an immediate and a one-month test (8 per condition); eighth-graders tutored interactively learned as much immediately and transferred better; a meta-analysis of dynamic assessment ranked explicit strategy training above contingent scaffolding for cognitive tasks. Evidence is almost all one-to-one.
  • Limits. Expertise reversal [~M]: guidance that helps novices (integrated explanations, worked examples, step-by-step guidance) can become redundant for more knowledgeable learners; one review, read from its abstract. Activation of topic knowledge [~M]: an early advance-organizer meta-analysis was favourable, while a randomized experiment with 88 primary pupils found activating topic knowledge did not improve comprehension and activating metacognitive knowledge did; a third study compared two amounts of activation with no control. So step 1 earns its place by revealing wrong ideas, not by priming. Whole-task practice [~M] is a training-design argument, not a study, that whole-task practice favours transfer of complex skills; it qualifies running this pattern as a separate lesson before a complex-skill is practised.

Do not rank this pattern against explain-then-practise or discovery lessons from these claims: none compares the sequence, and none measures how long each step should take. Do not read fluent recall of the definition, or correct answers given with prompts, as understanding. A precise abstention names the missing comparison (this five-step lesson against explanation plus independent practice, on an unseen prediction-and-reason item at a stated delay) and the local observation that would stand in for it.

The evidence by question

Further evidence, not yet read against this model

Claims this page cited before the 2026-10-05 rewrite and not used as core evidence above.

Illustrative design instance and observation record

Invented illustration, not a tested lesson. A Year 8 science class of 28 (adolescent, classroom) meets the idea that the seasons are caused by the tilt of the Earth's axis, over three 50-minute lessons and a check two weeks later. At the start of lesson 1 every pupil writes a prediction: is it summer when the Earth is closest to the Sun, and why? Nineteen say yes, because closer means hotter. The teacher names that idea, shows that the Earth is closest to the Sun in January, when it is winter in the northern hemisphere, and explains tilt with a lamp and a globe and a diagram of sunlight spread over a larger or smaller area. Pupils work two cases (Australia in December, the Arctic in June), writing "this happens because…" for each, and the teacher reads them as she circulates. The check is a new case: why is midday in summer hotter than early morning? Lesson 2 opens with two retrieval items, uses a table of day lengths and temperatures as a second representation, and compares the seasons with a case where distance does matter (a planet with a very elongated orbit). In lesson 3 pupils write their own case and explain it to a partner. One pupil says he wants to know why his grandmother in Brazil has Christmas in summer; the designer's objective is a correct prediction with a correct reason on an unseen case. They agree to use his case in lesson 3. The misconceptions claim covers refutation in science classes, not this sequence; the timings and the two-week check are local choices to observe.

Record: the idea and its known wrong versions → each learner's starting prediction and reason → explanation and representations used, and whether a wrong idea was named and refuted → cases worked, prompts given, help given → each learner's prediction and reason on the changed item → repair given and the re-check → comparison used in step 5 → candidate interpretations (idea grasped, tied to one representation, wrong idea suppressed but held, restating without a reason) and their basis → reobservation at the stated delay. Preserve the learner's stated purpose and any change in it. An agent may summarize this record but must not fabricate missing inputs or assign numerical probabilities of understanding without a calibrated model.

Elements and limits

Activation and prior knowledge activation, problem presentation, demonstration, worked examples, self-explanation, practice, feedback, formative assessment, comparing cases, non-examples and transfer tasks. Principles the pattern draws on: Activation, Direct Instruction, Guided Practice, Self-Explanation, Cognitive Disequilibrium, Analogical Reasoning, Creating Visual Representations, Immediate Feedback and Explicit Instruction: Vocabulary.

This pattern is scoped to building understanding of a concept or principle over one lesson or a short unit. Fluent procedures, complex skills practised as whole tasks, concepts best induced from examples, and open inquiry need other configurations (Direct Instruction, Merrill's First Principles, Concept Attainment). The policy supports observation and design reasoning; whether this sequence produces better prediction and explanation on unseen cases than explanation followed by independent practice, for which learners, at what time cost and at what delay, remains to be tested.

Claims

Examples

  • A concept lesson that starts with a familiar example, introduces a new model, and checks understanding through short application tasks.
  • A literacy lesson that connects prior experience to a new comprehension strategy and then practices it on a brief text.

Key Sources

  • Rosenshine, B. (2012). Principles of instruction: Research-based strategies that all teachers should know. American Educator, 36(1), 12-19, 39.
  • Pearson, P. D., & Gallagher, M. C. (1983). The instruction of reading comprehension. Contemporary Educational Psychology, 8(3), 317-344. https://doi.org/10.1016/0361-476X(83)90019-X
Page metadata
Field Value
type pattern
id develop-understanding
title Develop Understanding
description For a learner meeting a concept or principle, a lesson that elicits their existing ideas, explains the idea while refuting any wrong one, guides its use with self-explanation prompts, checks on a changed item and connects it by comparing cases is expected to improve explanation and prediction on new cases, though no claim tests the sequence as a whole; its components are tested separately.
status review
grain_size lesson
author Pearson
canonical True
generated by claude/unspecified on 2026-10-05
trust unverified — no one has confirmed this page's content

Frontmatter is the page's machine-readable record — see the Schema & Guide.