Skip to content

Guided Practice

Principle · All principles
Evidence · 13 claims (6 for, 7 mixed) · 23 studies (8 causal, 7 review, 3 theoretical, 2 quant-synthesis, 2 qualitative, 1 associational), q1–q4 · 1 of 23 report an effect size · 8 claims rest on one study

Conditional relationship

This page owns the practice phase between a model and independent work. For a learner who has seen an explanation or a worked model of a structured target (a procedure, concept or principle with a defined correct performance, or a component of a complex-skill), but whose own first attempts are slow, inaccurate or stall at a decision point, several practice attempts on comparable items with help and feedback still available are expected to produce more accurate unaided performance on that target (immediate-performance, near-transfer) than either no practice or practice left to the learner alone, where early errors can go unnoticed and be rehearsed. The support is reduced as the learner succeeds without it. Whether the gain holds at delayed-retention or far-transfer is a separate claim the evidence here does not settle.

The relationship is conditional on the learner's task-specific starting response (novice for this task; guidance can become redundant for an advanced learner), on whether someone or something can actually see each attempt and respond to it (most contingent-support evidence is one-to-one), on the success rate during practice, and on the horizon. No claim in the wiki compares guided practice with independent practice after the same model. The nearest evidence compares practice with no practice, contingent with fixed support, and interactive with explanatory tutoring.

How the converted siblings sit inside this: Direct Instruction holds the whole explain–model–guide–hand-over relationship and its comparison with discovery; guided practice is one of its phases, studied here on its own. Worked Examples holds what comes before the first attempt (studying a solution first). Scaffolding and Fading holds how support is raised, lowered and withdrawn; this page uses that model for the hand-over and does not repeat it. The Direct Instruction pattern and the Cognitive Apprenticeship pattern both carry a guided-practice step as a reusable policy; their branches are untested proposals.

Default design, while the relationship is untested

The earlier page's guidance, kept as a concrete default a designer can act on and revise. Each step is a design proposal; its evidence status is stated beside it.

  1. Model first, then practise at once. Start guided practice straight after learners have seen or heard what good performance looks like, on items close to the model (same structure, new surface). Untested here as a sequence; from the earlier page. What comes before the first attempt is modelled on Worked Examples. When re-modelling after errors, think aloud at the decision point the errors show (what you check, and why), so the learner hears the step they are skipping; untested proposal.
  2. Give several rounds, not one. Plan at least three or four supported attempts on comparable items per new step, so the learner can act on feedback from one attempt in the next. The earlier page asked for "multiple attempts"; the number is an untested proposal. Including practice at all, against a lesson without it, has one supporting experiment (computer-based, undergraduates; below).
  3. Watch each attempt and respond to it. Help where the attempt shows it is needed (a cue first, then a more specific hint, then a demonstration of the step), and less after a success. Evidence: contingent support beat fixed or no support in one-to-one tutoring of children on long division (below); untested at class scale.
  4. Correct errors during practice, not after it. Give feedback on each attempt before the next, so that an error is not practised again. Untested as a timing rule here; from the earlier page. A neighbouring finding, carried from retrieval testing: answering multiple-choice items without feedback can leave the chosen wrong answer to be reproduced later (below).
  5. Keep most attempts successful, but not trivially so. If many attempts fail even with the help available, the items are too far from the model: return to the model or a simpler item. If almost every attempt succeeds without any help, shorten guided practice and move on. No study here sets a success rate for skill practice, so set the threshold locally and record it; the only figure in the wiki, from a retrieval-practice moderator (below), is about recall and is not a rule for this.
  6. Hand over on unaided success, checked. Before removing support, ask the learner to do or explain the next step unaided; withdraw only after that succeeds on two or three consecutive items (an untested proposal). How to fade is modelled in Scaffolding and Fading.
  7. Reconnect parts to the whole. If the target was broken into parts, end with at least one guided attempt at the whole task before independent work. Untested; from the earlier page's caution against over-fragmentation.
  8. Include items where the step is not needed. Mix in items that do not call for the new step, so the learner practises deciding whether it applies, not only carrying it out; an error at a "hidden decision point" (the earlier page) shows up only when the decision is required. Untested proposal.
  9. Plan the withdrawal of support across sessions, not only within one. When the target is weeks away, a default arc: early sessions with the model visible and the teacher's cue–hint–demonstration ladder; middle sessions with the teacher's prompts replaced by a self-check the learner runs (a question card, an inverse check, a worked answer to compare); final sessions with mixed items in the form of the target assessment and help only on request. Open each session with two or three unaided items on new material, and move a learner to independent or extension work once they are accurate on two consecutive unaided rounds, so those already accurate are not held in guided practice (expertise reversal [~M]). The arc and its thresholds are untested proposals; how to fade is modelled in Scaffolding and Fading.

When a condition is not met: if no one can see attempts as they happen (a large class, self-paced online work), use items whose answers can be checked automatically, with worked feedback after each one, and a short unaided check before moving on; if learners already succeed unaided on the first item, skip to independent or whole-task practice.

Fitting the design to a situation

The three facts that most change the decision:

  • The learner's starting response on this task. A learner who cannot begin needs the model again and closer items; one who succeeds unaided on the first item needs little or no guided practice, and extra guidance can become redundant. If unstated, ask: can they do one item of this kind now, unaided, and explain a step?
  • Who can see each attempt, and how soon. The evidence for contingent help comes mostly from one-to-one tutoring; with one teacher and thirty learners, or no teacher, the feedback has to come from the items, peers or software. If unstated, ask: during practice, who will look at a learner's attempt before the next one, and how long after?
  • The capability and horizon that count. Accurate steps on items like the model, explaining why, solving changed problems, and doing it a week later need different practice and different checks. If unstated, ask: what must the learner do, unaided, with what aids, and when?
If the brief says… Then change… Basis
Learners are advanced on this task, or succeed on the first item unaided Drop most guided rounds; give one check item, then independent or whole-task practice with selective feedback Instructional guidance that helps novices can become redundant or counterproductive as expertise grows. [~M] (review of cognitive-load studies; gives no crossover point)
Learners are child (primary age), or adult / professionals Child: keep the cue–hint–demonstration ladder with one-to-one or small-group checking, and short rounds. Adult or professional: ask what they already do, and move guided items onto tasks from their own work as soon as the step is stable Child: Contingent scaffolding improves learning more than fixed or absent support. [+M] (fourth and fifth graders, long division, 8 per condition). Adult: untested proposal
Setting is online-self-paced, no teacher present Build practice items with automatic checking and feedback after each item into the lesson; do not cut them to save time; add an unaided check item before the next topic In one experiment with 256 undergraduates, computer-based lesson versions that included practice produced significantly higher posttest scores than versions without practice [+M] (undergraduates, one computer-based lesson; horizon not reported)
Practice items are multiple choice Give the correct answer and a short reason after every item, so a chosen wrong option is not kept Taking initial multiple-choice tests without feedback can lead students to later produce the incorrect lure answers they selected [+W] (college students, retrieval testing, reported second-hand; untested for skill practice)
classroom with one teacher and a large group Have every learner show each attempt at once (mini-whiteboards, cards, a quick poll) so the teacher can scan them all; give whole-class feedback on the commonest error, then go to those who still fail. With older learners, pairs can also check each other against a worked answer; with young children, or when one misconception is widespread, the teacher's scan is safer than peer checking. Collect one unaided item from everyone before independent work untested proposal; contingent-support evidence is one-to-one and its classroom-scale effect is not shown (Contingent scaffolding improves learning more than fixed or absent support. [~M])
Goal is a concept, principle or transfer, not only a procedure Replace some feedback-giving with prompts that ask the learner to explain or predict the next step; include one changed item in each round Human tutoring has a medium effect over no tutoring, and in one small study students learned as well when tutors only prompted them as when tutors also explained and gave feedback [~M] (eighth graders, 11 tutor–student pairs: prompting without feedback learned as well)
Time is one session only Plan the session as model, three or four guided items, one unaided check; record what the check shows and do not claim retention untested proposal
Many attempts fail with help, or nearly all succeed without it Failing: return to the model or simpler items. Succeeding unaided: shorten or end guided practice. Set the threshold locally and record it untested proposal; no claim sets a success rate for skill practice (a retrieval-practice moderator, Retrieval practice effects become more robust as initial retrieval success increases, especially above 75% [~M], is about recall)
The task has many steps or hidden decision points, or an error on the job is costly (safety, compliance, harm to a client) Plan more guided rounds per step; raise the release criterion (more consecutive unaided successes, on representative tasks); add a stage of unassisted work under supervision before full independence; define what "working alone" means (no supervision, or supervision with a way to escalate) untested proposal; from the earlier page ("multiple steps, hidden decision points, or a high likelihood of early error")
Stakes are high, or the assessment is unaided Judge readiness only on unaided items, never on assisted success during practice; keep a record of every hint given untested proposal
Instruction is in the learners' second language, or learners have low literacy Keep the language of practice items simpler than the target task; let learners show a step by doing it before explaining it; gloss terms the objective does not include untested; from the earlier page (adult literacy and numeracy settings)

Four of the eleven rows rest on a claim, and only one of those (computer-based practice with undergraduates) tests practice as such; the others carry a neighbouring finding (a cognitive-load review, retrieval testing, one-to-one tutoring) to this setting and say so. The age row rests on a claim for children only; the classroom and success-rate rows are untested proposals that cite a claim only for its limit; the rest are untested proposals.

Observation, state and explanation

What can be observed is an attempt under stated conditions: the item, the help available, the help requested or given (record its level), the answer, the time taken, and what the learner did with the feedback on the next item. "Has the routine", "is dependent on support" and "is rehearsing an error" are inferred states. Assisted success shows what the learner can do with this help; it does not show what they can do without it.

Response Plausible explanations Observation that could change the interpretation
Correct on guided items, wrong on the first unaided item The help carried the decision; the unaided item differs more than intended; the learner had it but lost it under time pressure or anxiety Give a matched unaided item with no time limit, then one with only a minimal cue; ask for the first step and why.
Makes the same error across several rounds Feedback was not seen or not understood; a misconception the feedback did not address; the learner is copying an earlier wrong attempt Ask the learner to say what the feedback meant, and give an item where the misconception and the correct rule predict different answers.
Asks for help on every item, including ones already done correctly Uncertainty about own accuracy; help is cheaper than effort; anxiety about being wrong in front of others Offer the item with help optional and private, and ask for a confidence rating before checking; compare accuracy with and without help taken.
Fast and accurate on guided items, cannot explain any step A procedure without its rationale; a language barrier to explaining; the items are too close to the model to need understanding Pose a changed item the memorised steps would not solve, and allow a non-verbal explanation (point, draw, act).
Accurate at the end of the session, much weaker a week later Little consolidation; guided items too close to the model; interference from later work Re-test with a new item at the delay and record all practice and exposure in between.

These rows are proposals, untested with learners as diagnostics. A probe is itself practice and changes the state it measures, so record each one, and keep the explanations that remain open rather than choosing the one whose name fits.

Evidence and qualifications

No claim in the wiki compares guided practice with independent practice following the same model, or varies the number of guided rounds. The claims below test neighbouring parts of the relationship.

  • In one experiment with 256 undergraduates, computer-based lesson versions that included practice produced significantly higher posttest scores than versions without practice [+M]: 256 undergraduate computer-literacy students, randomly assigned within pretest blocks to six versions of one computer-based lesson (randomized); the four versions that included practice scored significantly higher on the posttest than the two without it (means above 17 against below 15). It supports including a practice phase. The claim page's title says "practice with feedback", but its evidence entry does not describe the practice or its feedback, no effect size is printed, and the horizon is not reported. It does not compare guided with unguided practice. Not yet checked against its source.
  • Contingent scaffolding improves learning more than fixed or absent support. [+M]: fourth and fifth graders learning long division one-to-one under fully contingent support solved more problems than under fixed, partly contingent or no support, at an immediate and a one-month test (8 learners per condition, coded as a quasi-experiment). A meta-analysis of dynamic assessment ranked explicit strategy training above scaffolding and scaffolding above coaching. It supports responding to each attempt during practice; it does not show a classroom-scale effect, and contingency is reported to be scarce in whole-class teaching. Not settled: the abstracts could not confirm the entries.
  • Human tutoring has a medium effect over no tutoring, and in one small study students learned as well when tutors only prompted them as when tutors also explained and gave feedback [~M]: in tutoring of eighth graders on the circulatory system (11 tutor–student pairs per study), tutors told to withhold explanations and feedback and only prompt produced learning as good as ordinary tutoring; a review reports human tutoring at d = 0.79 and intelligent tutoring systems at d = 0.76 against no tutoring (abstract only). The claim's own Discussion says its evidence supports scaffolding more than feedback, so its title overstates the feedback half. It qualifies this page: the active ingredient of support during practice may be what the learner is prompted to construct, not the feedback given. Not settled: the abstracts could not confirm the entries.
  • Taking initial multiple-choice tests without feedback can lead students to later produce the incorrect lure answers they selected [+W]: a review chapter reports that students who answered multiple-choice items without feedback tended to reproduce the wrong options they had chosen on a later cued-recall test, and that college students answering SAT II items intruded more wrong answers on a short-answer test five minutes later than without a test. It supports the earlier page's worry that unguided practice can rehearse errors, but it is second-hand, about retrieval testing of facts rather than skill practice, and prints no effect size. One of two entries passes the judge.
  • Retrieval practice effects become more robust as initial retrieval success increases, especially above 75% [~M]: a review chapter reports, from a meta-analysis, that retrieval practice effects grow more robust as initial success rises, especially above 75%, while retrieval made too easy (more cues, massed immediately after study) yields smaller effects. Both entries pass the judge. It is the only figure in the wiki for a success rate during practice, and it comes from retrieval practice, not guided skill practice, so the default design does not take a threshold from it.
  • Instructional guidance that helps novices can become redundant or counterproductive as expertise grows. [~M]: a review of cognitive-load studies: guidance that helps novices can be redundant for learners who already hold the schema, and can depress their performance. It is the reason to shorten or end guided practice for learners who succeed unaided; it does not say when an individual crosses over. Not settled: the abstract could not confirm the entry.

Learners, tasks, settings and outcomes differ across these claims (undergraduates on a computer lesson, children with a tutor, college students on recall tests), so do not rank them by effect label. None measures delayed retention of a guided-practice gain against independent practice.

The evidence by question

Further evidence, not yet read against this model

Claims this page cited before it was rewritten as a conditional model. They are evidence about the relationship, but none has yet been re-read against the model above; each says how far its own sources have been checked.

Objective and learner-valued goal

The designer's objective is usually unaided performance on a representative task, with stated aids, at a stated horizon. Elicit separately what the learner wants from practice: to get today's worksheet done, to avoid being seen making mistakes, to be able to do the real task in their own life or job. Agreement and divergence are both observations. A learner who values finishing quickly may take every hint; one who values not looking stuck may refuse help they still need. Help-seeking during practice reflects the goal as well as the competence.

For example, an adult in a literacy programme may want to fill in a job application form on their own, while the designer's objective for the session is decoding multisyllable words in a graded reader. Guided practice on the reader serves the designer's objective; one guided attempt at an actual form, then an unaided one, tests the learner's. Record both, so that success on reader passages is not read as the capability the learner came for.

What would revise this model?

The expectation should weaken if comparisons of guided against independent practice after the same model, with comparable learners and aligned unaided outcomes, show no advantage, or show that a gain at the end of the session disappears at a delay. If prompting learners to construct the next step matches feedback-giving support, as one tutoring study suggests, the model should say "prompted practice" rather than "practice with feedback". If guided practice works no better at class scale, where few attempts are seen, than self-checked practice with automatic feedback, the contingent part is a one-to-one design preference only. If success rates during skill practice behave unlike those in retrieval practice, drop the three-in-four guide. Do not protect the model by calling every failure "too little practice" or "support removed too early" after the fact.

Assisted success during practice, unaided immediate performance, delayed retention, transfer and use in the setting the learner values are separate claims. The present evidence establishes no number of guided rounds, no success threshold for skill practice and no hand-over criterion.

Claims

  • Worked Examples — guided practice often follows worked examples so learners can apply a model immediately.
  • Scaffolding and Fading — guided practice is one of the main places where contingent support and fading are enacted.
  • Immediate Feedback — the quality of guided practice depends heavily on quick correction and coaching.
  • Chunking — breaking complex tasks into manageable parts often makes guided practice more workable for novices.

Examples

  • Teacher-led problem solving: Learners solve a near-transfer item while the instructor prompts, checks, and intervenes only where needed.
  • Supported writing conferences: Learners draft or revise with short targeted coaching on one or two moves at a time.
  • Guided reading or decoding practice: Learners rehearse a specific reading move with immediate cueing and correction.
  • Software practice with hints: A digital system provides step-level prompts that fade as accuracy improves.

Key Sources

  • National Research Council. (2012). Improving adult literacy instruction: Supporting learning and motivation. National Academies Press. https://doi.org/10.17226/13468
  • Numrich, C., & Kennedy, A. S. (2017). Providing guided practice in discourse synthesis. TESOL Journal, 8(1), 28-43. https://doi.org/10.1002/tesj.258
  • Massengill, D. (2004). The impact of using guided reading to teach low-literate adults. Journal of Adolescent & Adult Literacy, 47(7), 588-602.
  • Hua, Y., Morgan, B. S., Kaldenberg, E. R., & Goo, M. (2012). Cognitive strategy instruction for functional mathematical skill: Effects for young adults with intellectual disability. Education and Training in Autism and Developmental Disabilities, 47(3), 345-358. doi:10.1177/215416471204700309
  • Merrill, M. D. (2002). First principles of instruction. Educational Technology Research and Development, 50(3), 43-59. https://doi.org/10.1007/BF02505024
Page metadata
Field Value
type principle
id guided-practice
title Guided Practice
description For a learner who has seen a model but cannot yet perform a structured task alone, several practice attempts with support and feedback still available, reduced as unaided success appears, are expected to produce more accurate independent performance than no practice or unsupported practice, though no claim here tests guided against independent practice directly.
status review
moved_from pattern/merrills-first-principles-application
canonical True
generated by claude/unspecified on 2026-10-02
trust unverified — no one has confirmed this page's content
sources 4 — national-research-council-2012, numrich-2017, massengill-2004, merrill-2002

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