Skip to content

Deliberate Practice

Principle · All principles
Evidence · 17 claims (9 for, 8 mixed) · 31 studies (13 quant-synthesis, 7 review, 6 theoretical, 5 causal), q1–q4 · 11 of 31 report an effect size · 8 claims rest on one study

Conditional relationship

This page owns targeted refinement practice: what to do once a learner can already carry out a task, at least roughly and unaided, but has stopped improving on some part of it. For such a learner, working on one diagnosed component of the performance (not the whole task, and not whatever feels comfortable), at a level just beyond what they can do reliably, judged on every attempt against an explicit standard, with feedback before the next attempt and a deliberate change on that attempt, is expected to improve performance of the whole task (skill-performance, near-transfer) more than the same time spent repeating the whole task or practicing without a standard. The relationship is conditional on four things: an evaluable standard (the performance can be observed and judged against a stated criterion, and judged quickly enough to inform the next attempt); a learner beyond first exposure (an intermediate or advanced performer on this task; a novice needs a model and guided practice first); a source of feedback that can see each attempt; and sustained effort, which this kind of practice demands and which can collapse if progress is not visible.

No claim in the wiki tests this relationship in schooling. The nearest causal evidence is a meta-analysis of 14 studies in which simulation-based medical education built around deliberate practice (repetition, mastery standards, immediate feedback) beat traditional clinical education on procedural and psychomotor skills; it compares a bundle with ordinary training, not targeted practice with whole-task repetition at equal time. One self-selected comparison of real estate licensing students points the same way. The rest of the deliberate-practice evidence is correlational: accumulated hours of deliberate practice are associated with performance, strongly in games, music and sports and weakly in education and the professions, and practice hours vary so widely among experts that practice cannot be the only factor. So the page offers a default design, labelled step by step, not a tested model.

How the converted siblings sit beside this: Guided Practice owns the phase straight after a model, when a novice cannot yet perform unaided; this page starts where that one hands over. Goal Setting and Monitoring owns what the goal should be (specific, challenging, process or outcome) and how progress is shown; this page uses it for the component goal and does not repeat it. Immediate Feedback owns when corrective feedback should arrive. Retrieval Practice owns recall practice for facts and prose, which is a different kind of practice from refining a performance. Mastery Learning owns the gate that decides when a learner moves on to the next unit; Scaffolding and Fading owns how support is withdrawn.

Default design, while the relationship is untested

Each step is a design proposal; its evidence status is stated beside it.

  1. Start from a scored whole-task attempt. Have the learner perform the whole task once, unaided, and score it against a rubric of two to four named dimensions (in writing: thesis clarity, use of evidence, sentence control; in a procedure: each critical step and its timing). Keep this as the baseline. Untested proposal; the earlier page required "an evaluable standard".
  2. Choose one component to work on. Pick the dimension that fails most often across two or three attempts, not the one the learner likes, and confirm it with a short isolated probe (one paragraph, one step, one passage). Work on one component per round. Untested proposal; from the earlier page ("targeted revision rounds", "error-focused rehearsal").
  3. State the goal as what to do, not only the score to reach. Write the component goal as a specific action ("open each paragraph with the claim it supports"), set just beyond what the learner does reliably now. For a learner still near the start, keep the goal on the process; once the process is under control, add an outcome target. Evidence: process goals beat outcome goals for novices [~M] (high-school writing revision; the authors also report that shifting to outcome goals once the skill is mastered can help); specific, difficult goals beat vague ones [+M] (a narrative review of work and laboratory tasks); goal setting improves performance [~S] (two meta-analyses: specific, difficult goals raised task performance, but less on complex tasks, d = .76 on easy tasks to d = .42 on the most complex, and setting a goal had a small effect on behaviour change, d = .34; work, laboratory and health tasks, not learning outcomes), so expect a smaller push from the goal when the component sits inside a complex performance. The goal itself is modelled in Goal Setting and Monitoring.
  4. Build short drills that isolate the component at the edge of reliable performance. Make items that call on the component alone, at a difficulty where the learner succeeds on some attempts but not all, with full effort. Proposal: if fewer than about half the attempts in a round meet the standard after two rounds, make the items simpler or show a model again; if nearly all meet it, make them harder or move on. These thresholds are untested; no claim sets a success rate for refining a skill.
  5. Judge every attempt against the standard, and give feedback about the task before the next attempt. Say which part of the attempt met the standard, which did not, and what to change, not how good the learner is. Evidence: feedback improves learning [+S] on average, most when it carries more information, and in one meta-analysis more than a third of feedback interventions made performance worse, more so the more they drew attention to the self. When the feedback should arrive is modelled in Immediate Feedback.
  6. Make the next attempt different. Before each retry, the learner says (or writes) what they will change. Give three to five attempts per round so feedback from one can be used in the next. The number is an untested proposal; the earlier page required "repeated attempts with adjustment".
  7. Have the learner keep an error log. After each attempt the learner records which criterion they missed and what they tried; at the end of the session they read the log for a pattern. Evidence: self-monitoring improves self-regulation [+M] (a review and a theoretical synthesis; not tested within deliberate practice).
  8. Put the component back into the whole task. End every session with one whole-task attempt scored on the same rubric as the baseline, so a gain in the drill is checked where it matters. Untested; from the earlier page ("then reintegrate it into whole performance").
  9. Move on by a standard, not by time spent. A component is done when it meets the standard in the whole-task attempt in two consecutive sessions, unaided, and again at a check about a week later. Mastery standards rather than fixed time are part of the design the clinical-simulation evidence pooled (simulation-based education with deliberate practice [+M], health professions, procedural skills); the two-session and one-week thresholds are untested proposals.
  10. Plan the arc across sessions. A default for a skill worked on over several weeks (all quantities are untested proposals): sessions of 20 to 40 minutes of focused practice, three or four a week, each opening with a two-minute re-attempt of a component already finished. Weeks 1 to 2: the coach chooses the component, models it, and gives the feedback; the learner keeps the log. Weeks 3 to 4: the learner scores their own attempts against the rubric and annotated exemplars, the coach checks one attempt in three and the learner proposes the next component from the log. Week 5 on: the learner chooses the component, sets the standard with the coach, and asks for feedback when the log shows a plateau. Revisit finished components at intervals rather than dropping them (spaced practice improves retention [+S]: mostly retention of studied verbal material; its one meta-analysis of task practice, including motor skills such as typing, found spaced beat massed practice on performance, d = 0.46, with the effect depending on task type and the interval, smaller for complex tasks, and longer lags reported to reduce it for skill tasks such as typing, gymnastics and music). So keep the gaps between revisits of a refined skill short at first; growing them is carried from verbal recall, where expanding gaps were no better than even ones (spaced retrieval [~S]), and is an untested proposal for a skill. The earlier page noted that full-concentration practice is effortful; the deliberate-practice claim page puts diminishing returns beyond about an hour at weak evidence, so a session longer than that should be split.

When a condition is not met: if no one can judge attempts as they happen, use items with a worked answer or an annotated exemplar the learner compares against, and keep the coach's role to checking the self-scoring; if the learner cannot yet perform the whole task unaided, use Guided Practice first; if no standard can be stated, the first job is to build one (exemplars at several levels), not to practice.

Fitting the design to a situation

The three facts that most change the decision:

  • Whether the performance can be judged against a standard, and how fast. Deliberate practice is easiest to design where each attempt can be scored at once (a clinical procedure on a simulator, a musical passage, a maths step) and hardest where quality is diffuse and feedback slow (an essay's argument, teaching, management). If unstated, ask: what does a good attempt look like, who can say whether this attempt meets it, and how soon?
  • Where the learner is on the whole task. A learner who cannot yet do the task unaided needs a model and guided practice, and isolating components early can overload them; one who performs but has plateaued is the case this page is for. If unstated, ask: can they do the whole task once now, unaided, and which part goes wrong most often?
  • Who supplies the feedback and the effort. The evidence pools designs with a coach or simulator judging every attempt; with one teacher and thirty learners, or none, the judging moves to rubrics, peers or software, and the learner's willingness to keep at effortful practice becomes the limit. If unstated, ask: who will look at each attempt, and what does the learner want from getting better at this?
If the brief says… Then change… Basis
Learners are novice on this task, at first exposure Do not isolate components yet: model the whole task, use guided practice with support, and set process goals; start targeted refinement only once the learner can perform unaided Instructional guidance that helps novices can become redundant or counterproductive as expertise grows. [~M] (review of cognitive-load studies, carried to skill refinement); process goals for novices [~M]
Learners are advanced and have plateaued Let the learner pick the component from their own log, raise the standard, use outcome targets beside process goals, and give feedback mainly when the log shows a plateau Process goals lead to better skill acquisition for novices than outcome goals. [~M] (reports that a shift to outcome goals after mastery can help; high-school writing)
Learners are child (primary age) The teacher chooses the component and writes the goal in the child's words; drills of five to ten minutes; a visible chart of attempts that met the standard; whole-task attempt at the end untested proposal
Goal is a diffuse performance (complex-skill such as essay writing, teaching, leading a meeting) Build the standard first: three or four annotated exemplars at different levels and a short rubric; expect smaller and slower gains than in sport or music, and check gains on new whole tasks; expect the component goal to push less than on a simple task Deliberate practice and performance across domains [~M] (correlational meta-analysis: practice hours explain far more of performance in games, music and sports than in education or the professions; not a test of a design); Goal setting improves performance [~S] (goal effects smaller on complex work and laboratory tasks)
Goal is a procedure or psychomotor skill in workplace-clinical training Use a simulator or part-task trainer with a mastery standard per step, immediate expert or device feedback, repetition until the standard is met and rising difficulty; add booster sessions later Simulation-based education with deliberate practice improves clinical outcomes [+M] (14 studies, health professions, procedural skills, against traditional clinical education)
Goal is recall of facts or vocabulary (verbal-association) Use spaced retrieval practice with corrective feedback instead of component drills Retrieval practice and spaced retrieval improve achievement and retention by medium amounts, while interleaving helps category learning only for some materials and blocking beats it for words [~S] (spaced retrieval beats massed; interleaving helps some materials and not words); see Retrieval Practice
Setting is online-self-paced, no coach present Give items with automatic checking or a worked answer to compare; have the learner score against annotated exemplars and keep an error log; require one whole-task submission per session reviewed by a person or a rubric-based check Self-regulated learning models treat self-monitoring as a core process through which learners notice gaps and adjust strategy [+M] (reviews; untested within deliberate practice or online)
classroom with one teacher and a large group Work on the component most of the class misses; pairs score each other's attempts on that one rubric line; the teacher scans every whole-task attempt at the end and gives whole-class feedback on the commonest error untested proposal
Time is one session only One component, three rounds of three to five attempts, then a scored whole-task attempt; report the change against the baseline as immediate only, not as retained untested proposal
Stakes are high (licensing or certification exam) Choose components from an item analysis of practice exams, practice in the exam's format with feedback on each item, and run timed whole exams at intervals Deliberate practice interventions produce higher real estate licensing exam pass rates [+W] (one dissertation, two self-selected groups, no effect size)
Learners are giving up, or the practice feels punitive Shorten sessions, lower the step to the next standard, show the log of attempts that met it, and let the learner choose which component to work on next untested; a practitioner assertion that over-challenge with too little support leads to giving up (Imbalanced challenge and support causes novice library learners to give up or remain dependent [~W], no data)
Feedback is mostly praise, grades or comparison with others Replace it with feedback about the attempt against the standard and the change to make next Feedback improves learning by a medium amount on average, but its effect varies widely with the information it carries, and a substantial share of feedback effects are negative [~S] (more than a third of feedback interventions reduced performance, more so when they drew attention to the self; work and laboratory as well as education)

Six of the twelve rows rest on a claim. Only the clinical-simulation row rests on a comparison of a deliberate-practice design with ordinary training, and that is in health professions; the others carry a neighbouring finding (goal setting, feedback, self-monitoring, expertise reversal, retrieval practice) or a correlational one and say so. The child, classroom, single-session and motivation rows are untested proposals.

Observation, state and explanation

What can be observed is an attempt under stated conditions: the task or drill item, the rubric line it was scored on, the score, the help or feedback given, what the learner said they would change, and whether the next attempt changed. "Has plateaued", "has fixed the component" and "is not trying" are inferred states. A drill score shows what the learner does on the isolated component; it does not show what they do in the whole task, and a coach's judgement and the learner's own may differ.

Response Plausible explanations Observation that could change the interpretation
Drill scores rise, whole-task scores on that component do not The drill isolated the component too far from its use; attention in the whole task goes elsewhere; the rubric scores the two differently Score the component in a whole-task attempt with a reminder to attend to it, and one without; compare the rubric line used in drill and in whole task.
Same error on attempt after attempt The feedback was not understood; the learner cannot see the difference between their attempt and the standard; the component is too far beyond current performance Ask the learner to point to the difference between their attempt and an exemplar; give a simpler item on the same component; check what they said they would change.
Plateau after early gains The standard has been met and the component is done; the drill has become too easy; fatigue or loss of effort; a different component now limits the whole task Re-score the whole task against the baseline rubric; raise the drill difficulty for one round; look in the log for a new recurring error.
Learner's self-scores differ from the coach's The learner does not yet hold the standard; the rubric line is ambiguous; the learner scores effort rather than result Have learner and coach score the same two exemplars independently and compare; rewrite the rubric line with an example at each level.
Effort drops, sessions skipped The gain is not visible to the learner; the component chosen is not one the learner cares about; the difficulty is too high Show the log of attempts that met the standard; ask what the learner wants to be able to do; offer a choice of next component and record what they choose.

These rows are proposals, untested with learners as diagnostics. A probe is itself a practice attempt 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 targeted component practice with whole-task repetition at equal time, varies the difficulty of practice relative to the learner's level, or tests the default design above. The claims below test neighbouring parts of it.

  • Simulation Based Education With Deliberate Practice Improves Clinical Outcomes [+M]: a meta-analysis of 14 studies (mostly randomized trials and pre/post comparisons) found simulation-based medical education with deliberate practice ahead of traditional clinical education on clinical skill acquisition (d = 0.71, 95% CI 0.65–0.76). The skills were procedural and psychomotor (resuscitation, laparoscopic and central-line procedures), the learners health professionals, and the comparison a bundle (simulator, mastery standards, feedback, repetition) against ordinary training, so it does not isolate the deliberate-practice features. One study (q3, not pre-registered); the claim page reports no horizon and calls patient-level outcomes less established.
  • Simulation Based Education Improves Outcomes [~M]: a meta-analysis of 609 studies of technology-enhanced simulation for health professionals found large effects against no intervention, and no consistent interaction with design features including feedback, mastery learning, distributed practice and repetitive practice; heterogeneity was large. It qualifies the reading that those features are the active ingredient in the previous claim. Against no intervention, not against an equal-time alternative.
  • Accumulated deliberate practice is associated with performance, explaining a domain-dependent share of its variance: about a quarter in games, 4% in education and under 1% in professions [~M]: one meta-analysis (Macnamara et al. 2014) of individual differences in accumulated deliberate practice found it explained 26% of performance variance in games, 21% in music, 18% in sports, 4% in education and under 1% in the professions. It is correlational: it tests whether people who practiced more perform better, not whether a design of practice causes a gain. The claim's title was narrowed to this correlational reading on 2026-10-05, and its second entry (Wedlock & Binnie 2025) reports the education figure second-hand. It is the reason the page expects smaller gains where the standard is diffuse.
  • Deliberate practice research is largely correlational with large unexplained variability in accumulated practice hours [~M]: a review reports very large standard deviations in the hours athletes accumulate, that the research is mostly retrospective and correlational, and that retrospective data may describe social background rather than causes. It limits any claim that more deliberate practice is the route to expertise.
  • Amount of accumulated deliberate practice distinguishes more from less accomplished musicians [+W]: the Berlin music academy study as reported second-hand in a review: by age 20 the best musicians had accumulated more than 10,000 hours of practice, more than two less accomplished expert groups and amateur pianists. Retrospective and correlational, in the domain where the meta-analysis above finds practice explains most; it says nothing about how to design a session.
  • Process goals lead to better skill acquisition for novices than outcome goals. [~M]: high-school students learning writing revision with process goals outperformed those given outcome goals, and a shift to outcome goals once the skill was mastered was reported to help; a theoretical chapter synthesizes similar results, including dart throwing. No effect size or sample recorded. It supports stating the component goal as an action, and changing the goal as the learner advances.
  • Feedback improves learning by a medium amount on average, but its effect varies widely with the information it carries, and a substantial share of feedback effects are negative [+S]: two meta-analyses: feedback raised learning on average (d = 0.48 over 435 studies, larger for high-information feedback, d = 0.99, than for reinforcement or punishment, d = 0.24) and performance on average (d = .41), but more than a third of interventions reduced performance, more so as feedback drew attention to the self. It supports feedback on every attempt and qualifies its form; neither synthesis is about deliberate practice as such.
  • Instructional guidance that helps novices can become redundant or counterproductive as expertise grows. [~M]: a review of cognitive-load studies. Carried here in the other direction: the decomposed, coach-directed practice suited to an advancing performer is not where a novice should start, and the coach's direction can be withdrawn as the learner's own standard forms. It does not give a crossover point.
  • Goal setting improves performance [~S]: two meta-analyses (125 goal-setting studies; 141 papers of randomized trials). Specific, difficult goals raised work and laboratory task performance, by less on complex tasks (d = .76 to d = .42), and goal setting had a small effect on behaviour change (d = .34). It qualifies step 3: the component goal is expected to help, by less as the performance it sits in grows complex; neither synthesis measured learning outcomes.
  • Spaced Practice Improves Retention [+S]: mostly verbal recall, but one meta-analysis of 63 studies of task practice, including motor skills, found spaced beat massed practice (d = 0.46), moderated by task type and the interval, with larger effects in less rigorous studies. It supports spreading practice of a component across sessions (step 10); it does not say how long the gaps should be for a refined skill.

Learners, tasks and outcomes differ across these claims (health professionals on simulators, high-school writers, athletes and musicians studied retrospectively), so do not rank them by effect size. None measures delayed retention or transfer of a targeted-practice gain in a school or university course.

The evidence by question

Further evidence, not yet read against this model

Re-read against the model on 2026-10-15; these two do not bear on its relationship, a step or a situation row.

Objective and learner-valued goal

The designer's objective is usually a whole-task performance that meets a stated standard, unaided, on new instances of the task, at a stated horizon. Elicit separately what the learner wants from getting better: to pass a licensing exam, to play a piece at a recital, to stop being corrected by a supervisor, to write an essay they are proud of. Agreement and divergence are both observations. A learner who values the exam may want only the components the exam scores; one who values a performance may resist drills that feel far from it. Deliberate practice asks for sustained effort, so whether the learner cares about the component being practiced is part of whether the design will be carried out at all.

For example, a nursing student's objective in a skills course may be meeting the checklist standard for a sterile dressing change; what the student values may be not being the one who contaminates the field in front of a patient. Practicing the glove step to the checklist standard serves both, but only a whole-task attempt under observation tests the second. Record both, so that a checklist pass is not read as the confidence the student came for.

What would revise this model?

The expectation should weaken if comparisons of targeted component practice against whole-task repetition at equal time, with comparable learners and an aligned whole-task outcome, show no advantage, or an advantage that disappears at a delay or on new tasks. If the clinical-simulation effect turns out to come from the simulator or the extra practice time rather than from the standard, feedback and repetition, the deliberate-practice features are not the active ingredient. If gains in education and the professions stay as small as the correlational evidence suggests even under well-built standards, the page's scope should narrow to well-defined skills with fast feedback. If practice set just beyond reliable performance works no better than easier practice with more successful attempts, drop that condition. Do not protect the model by calling every failure "not really deliberate" after the fact.

Drill performance, whole-task performance at the end of a session, retained performance, transfer to new tasks, and use where the learner wants it are separate claims. The present evidence establishes no session length, no number of attempts, no success rate during practice and no criterion for moving on.

  • Immediate Feedback — needed so learners can adjust performance quickly
  • Mastery Learning — often supplies the progression logic around deliberate practice
  • Scaffolding — supports can help learners practice just beyond current ability without losing task integrity
  • Guided Practice — the supported practice that comes before targeted refinement, for a learner who cannot yet perform unaided
  • Goal Setting and Monitoring — what the component goal should be and how progress against it is shown
  • Retrieval Practice — recall practice for facts and prose, a different kind of practice from refining a performance

Examples

Illustrative

Mastery Learning — Learners repeat focused practice-and-feedback cycles until a defined performance standard is met.

Targeted revision rounds — Writers practice only thesis clarity, evidence integration, or sentence control in one cycle rather than vaguely “improving the paper.”

Error-focused rehearsal — In music, sport, or math, learners isolate a recurring weakness, practice it with immediate feedback, and then reintegrate it into whole performance.

Simulation with a mastery standard — Health-professions learners rehearse a procedure on a simulator with immediate feedback until each step meets a checklist standard.

See also the strategies Deliberate Practice and Individual Deliberate Practice, and the theory page Deliberate practice framework.

Key Sources

  • Ericsson, K. A., Krampe, R. T., & Tesch-Romer, C. (1993). The role of deliberate practice in the acquisition of expert performance. Psychological Review, 100(3), 363-406. https://doi.org/10.1037/0033-295X.100.3.363
  • Ericsson, K. A. (2006). The influence of experience and deliberate practice on the development of superior expert performance. In K. A. Ericsson et al. (Eds.), The Cambridge handbook of expertise and expert performance. Cambridge University Press.
  • Do, P. D. (2024). Deliberate practice: An effective way to raise the real estate licensing passing rate. Bradley University. https://scholar.google.com/scholar?q=Deliberate+Practice+Real+Estate+Licensing+Passing+Rate
Page metadata
Field Value
type principle
id deliberate-practice
aliases structured-deliberate-practice-for-licensing-review
title Deliberate Practice
description For a learner who can already perform a task but has stopped improving on part of it, repeated attempts on that one diagnosed component, set just beyond reliable performance, judged against an explicit standard with feedback and a deliberate change on each next attempt, are expected to improve whole-task performance more than repeating the whole task; no claim tests this in schooling, and the nearest causal evidence is simulation training with deliberate practice in the health professions.
status review
canonical True
generated by claude/unspecified on 2026-10-15
trust unverified — no one has confirmed this page's content
sources 2 — ericsson-1993, do-2024

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