Skip to content

Procedural Learning

Principle · All principles
Evidence · 29 claims (16 for, 12 mixed, 1 against) · 46 studies (23 causal, 8 review, 7 theoretical, 6 quant-synthesis, 1 qualitative, 1 design), q1–q4 · 4 of 46 report an effect size · 19 claims rest on one study

Conditional relationship

This page owns the whole arc of learning a procedure, from a learner's first sight of it to fluent, retained and flexible unaided use, and three decisions along that arc that the neighbouring pages do not own: whether to teach the procedure whole or in parts; whether and when to tie the steps to their reasons and conditions of use; and how much practice to give after first accuracy, and how to spread and mix it across sessions. For a learner whose first response on a task shows they cannot yet carry out the procedure unaided (a novice for this task, on a procedure or a component of a complex-skill with a defined correct performance), a sequence of a visible model, several supported attempts with feedback, support withdrawn as attempts succeed, then continued unaided practice mixed with earlier procedures and spread over sessions, is expected to produce more accurate, faster and more durable unaided performance (immediate-performance, skill-performance, delayed-retention) than explanation alone or an equal amount of unsupported problem solving. Linking the steps to when and why they apply is expected to matter most for near-transfer and far-transfer, and less for accuracy on items like those practised.

The relationship is conditional on the learner's task-specific starting response (guidance that helps a novice can become redundant for an advanced learner), on the procedure's element interactivity (high: many steps that must be held and coordinated at once; low: steps that can be learned one at a time), on whether someone or something can see each attempt and respond, and on the horizon at which the procedure must work.

No claim in the wiki tests this arc as a whole, and none compares the full sequence with explanation alone. The claims below test single links in it: practice against no practice in one computer-based lesson; worked examples against problem solving for novices; contingent against fixed or absent support in one-to-one tutoring; process against outcome goals during early skill learning; isolated parts before the whole for novices on complex material; mixed against blocked practice of mathematics problem types; and side-by-side comparison of solution methods. Joining those links into one sequence is a design argument, not a finding. So the page offers a default design, labelled step by step.

How the converted neighbours sit inside this arc: Modeling owns making the performer's decisions visible before the first attempt, and Worked Examples owns studying a written solution first. Guided Practice owns the supported attempts straight after the model, and Scaffolding and Fading owns how support is raised, lowered and withdrawn. Deliberate Practice owns refining one component once the learner can perform the whole roughly but has stopped improving. Spaced Learning owns the spacing schedule in general, and Goal Setting and Monitoring owns what goal a learner works to. Direct Instruction holds the explain–model–guide–hand-over sequence and its comparison with discovery. The 4C/ID pattern is a reusable whole-task policy that includes just-in-time procedural information and part-task practice; its branches are untested proposals. This page links those pieces into one arc for a procedure and does not repeat their models.

Default design, while the relationship is untested

The earlier page's guidance (model, coached repetition, timely feedback, adaptive support, fading, conceptual grounding, staging of complex routines), kept as a concrete default a designer can act on and revise. Each step says what supports it.

  1. Write the procedure down as steps and decision points, and check where the learner starts. List the steps, the points where the performer must decide something (which method, whether a step applies, when to stop), and the conditions in which the procedure is used. Before teaching, give one representative item unaided and ask for the first step. A learner who completes it accurately skips to step 7. Untested proposal; from the earlier page's requirement of "a target procedure or routine".
  2. Decide whole or parts. If the procedure has many steps that must be coordinated at once and the learner cannot yet hold them (high element interactivity), teach the hardest two or three steps in isolation first, then join them within the same session or the next. If the steps can be learned one at a time, or the learner already has some of them, teach the whole sequence on a simple version of the task. Isolated parts first: Part-task practice reduces cognitive load for absolute novices during initial skill acquisition. [+M] (novices, complex high-interactivity material). Joining back to the whole: A training-design argument, not tested by any study recorded here, holds that practising whole complex tasks improves transfer of complex skills [~W] (a design argument, no experiment). The join point is an untested proposal.
  3. Show it worked out, one or two examples per new step. Demonstrate the procedure or give a worked example of it, saying what you check at each decision point and why, before the learner's first attempt; keep spoken explanation short until the learner can carry out the sequence (the earlier page's warning against over-explaining). Frame the learner's goal as carrying out the steps well, not as getting the final answer or score. Worked examples reduce unnecessary search for novices. [+M] (algebra, examples against the same time solving problems); Process goals lead to better skill acquisition for novices than outcome goals. [+M] (high-school students learning writing revision). The number of examples is an untested proposal; how to model is on Modeling.
  4. Give three or four supported attempts per new step, with feedback before the next attempt. Watch each attempt and help where it shows help is needed: a cue first, then a more specific hint, then a demonstration of the step; less help after a success. Correct an error before the next attempt so that it is not rehearsed. Including practice at all: In one experiment with 256 undergraduates, computer-based lesson versions that included practice produced significantly higher posttest scores than versions without practice [+M]. Adapting the help to each attempt: Contingent scaffolding improves learning more than fixed or absent support. [+M] (one-to-one tutoring of children; untested at class scale). The number of attempts is an untested proposal; the full routine is on Guided Practice.
  5. Withdraw support on unaided success. Remove the model and the hints once the learner completes two consecutive items unaided (an untested threshold), keeping the cue–hint–demonstration ladder available on request. How to fade is on Scaffolding and Fading.
  6. Tie the steps to their reasons and their conditions of use, once the steps are accurate. When a learner can carry out the steps, ask them to say why a step works and when the procedure applies, and set two correct methods for the same item side by side to compare (which is shorter here, and when would you use the other?). Include items where the procedure does not apply. Comparison of methods: 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] (seventh graders comparing equation-solving methods). Conditions of use: Strategy Instruction Needs Conditional Knowledge [~M] (reading strategies, one quasi-experiment and a theoretical paper; its own Discussion warns that long when-and-why discussion can overload novices before the steps are fluent). Placing this after the steps are accurate, not before, is an untested proposal.
  7. Keep practising after first accuracy, mixed with earlier procedures. Once each procedure is accurate on its own, give short practice rounds (5–10 items, 10–15 minutes) in which items of the new type are shuffled with earlier types the learner could confuse it with, so the learner must first choose the procedure. Expect more errors during mixed practice than during blocked practice; do not read them as failure. Interleaving Improves Discrimination [+M] (one experiment shuffling mathematics problem types, better on a delayed test and worse during practice; a meta-analysis finds the benefit largest when the types are easily confused). Round length and size are untested proposals.
  8. Spread the practice over sessions. A default arc for a procedure needed weeks later: session 1 (about 30–40 minutes) for steps 2–5 on the new procedure; sessions 2 to 4, two to four days apart, each opening with three to five unaided items on the procedure before any new teaching, then mixed rounds; later reviews at growing gaps (about one week, then three to four weeks). Distributed Practice Improves Retention [+M] (carried from verbal recall and fact learning to procedures; it gives no schedule for skills, and its optimal-gap figures are for facts). The arc and its gaps are untested proposals; spacing in general is on Spaced Learning.
  9. Switch the learner's goal from process to outcome once the steps are stable, for example from "carry out each step" to "finish five items correctly in ten minutes", if speed matters in the setting where the procedure will be used. Process goals lead to better skill acquisition for novices than outcome goals. [+M] (the study's results suggested a shift to outcome goals after mastery can help). The timing is an untested proposal.
  10. Move on, or hand over to refinement. Count the procedure learned when the learner is accurate unaided on mixed items at least two days after the last practice, on items in the form they will meet in use. A learner who is accurate but has stopped improving on one part goes to Deliberate Practice; a learner accurate on items like the practised ones but not on changed ones goes back to step 6. Untested proposal.

How the source and amount of support change from early to late sessions. Early: the teacher's or a worked example's model is visible, one procedure at a time, help after every attempt. Middle: the model is removed, the learner checks their own work against a worked answer or an inverse check, items are mixed with earlier types. Late: mixed items in the form of the real task, at gaps of a week or more, help only on request, and a goal of speed or finished output where speed matters.

Fitting the design to a situation

The three facts that most change the decision:

  • The learner's starting response on this procedure. A learner who cannot begin needs the model, parts and supported attempts; one who completes a representative item unaided needs little more than mixed and spaced practice, and extra guidance can become redundant. If unstated, ask: can they do one item of this kind now, unaided, and name the first step?
  • What kind of procedure it is. A fixed sequence that must run the same way every time (a dosage calculation, a lab protocol, a safety check) needs accuracy then speed; a procedure full of decisions (choosing a method, debugging, diagnosing) needs the decisions practised and the conditions of use taught. A procedure whose steps interact heavily may need parts first. If unstated, ask: does the procedure always run the same way, and where must the performer choose?
  • When and how it must work. Accurate on today's worksheet, fast enough in a real job, still accurate in a month, or adapted to a case not seen before each need different practice and a different check. If unstated, ask: what must the learner do, unaided, with what aids, how fast, and when?
If the brief says… Then change… Basis
Learners are advanced on this procedure, or complete the first item unaided Skip modelling and guided attempts; give mixed, spaced practice on full tasks and selective feedback, and refine weak components Instructional guidance that helps novices can become redundant or counterproductive as expertise grows. [~M] (cognitive-load studies; gives no crossover point)
The procedure has many interacting steps and learners are novice Teach the hardest steps in isolation for one or two short rounds, then join them on a simple whole task in the same or next session; never leave the parts unjoined Part-task practice reduces cognitive load for absolute novices during initial skill acquisition. [+M]; joining: whole-task argument [~W] (no experiment)
The brief wants a whole-task performance kept whole (a full clinical procedure, a complete repair, a full presentation) Keep every practice attempt a whole task; simplify the task (fewer variables, an easier case) rather than splitting it, and put the support around the whole; isolate a step only for a few minutes when one step keeps failing untested proposal; whole-task argument [~W]
Learners are child (primary age) Shorter rounds (3–5 items), concrete materials or a physical model of each step, one-to-one or small-group checking during supported attempts Contingent scaffolding improves learning more than fixed or absent support. [+M] (fourth and fifth graders, long division, one-to-one); round size untested
adult learners in a work setting, using the procedure on the job Ask what they already do; use items from their own work as soon as the steps are stable; give a job aid laid out in the workplace's own categories and step names, used during practice and kept afterwards; check on a real or realistic task untested proposal
The brief names its own checklist, job aid, rubric or routine for the procedure Keep it and its categories as given; teach and practise with it, and change only what the evidence speaks to: one model of each decision point, feedback before the next attempt, a delayed unaided check untested proposal; feedback before the next attempt from contingent scaffolding [+M] (one-to-one)
Goal includes transfer to changed problems, or understanding why, not only accuracy After accurate steps, add side-by-side comparison of two methods, items where the procedure does not apply, and "why does this step work?" prompts; check with one changed item per round Comparing cases side by side… [+M] (raised procedural flexibility, not conceptual knowledge, in one algebra study); Strategy Instruction Needs Conditional Knowledge [~M]
Learners can already run the procedure but cannot say what it means (e.g. can integrate but cannot explain the integral) Do not add more of the same practice; pose items the memorised steps will not solve and ask learners to explain or draw what each step does Students can apply procedural knowledge to solve integration problems but show limited understanding of basic integration concepts [~W] (calculus, reported second-hand)
Several similar procedures are taught in one unit (e.g. problem types, grammar forms, drug calculations) Once each is accurate alone, practise them shuffled so the learner must choose; warn learners that mixed practice feels harder and is slower Interleaving Improves Discrimination [+M] (mathematics problem types; benefit largest when types are confusable)
online-self-paced, no teacher present, or phone-only Build the worked example and five to eight practice items with automatic checking and a worked answer after each into the lesson; one step per screen on a phone; an unaided check item before the next procedure; scheduled review prompts two days and one week later practice included in computer-based lessons [+M] (undergraduates, one lesson; horizon not reported); schedule untested
Mixed modes: some learners in the room, some joining remotely Show the model on a shared screen or camera pointed at the hands or the work, not the whole room; have every learner, remote or present, submit each attempt the same way (photo, poll, shared document) so all are seen before feedback untested proposal
classroom, one teacher and a large group Every learner shows each attempt at once (mini-whiteboards, cards, a poll); whole-class feedback on the commonest error, then the teacher goes to those still failing; pairs check each other against a worked answer for routine steps untested proposal; contingent-support evidence is one-to-one (contingent scaffolding [~M])
Scarce facilitator time (volunteers, a busy shift, a tutor checking in briefly) Spend the facilitator's minutes on the first one or two supported attempts and the unaided check; give everything between to self-checking items with worked answers or a job aid; the facilitator looks only at the step the learner flags untested proposal
Time is one session only Model, three or four supported attempts, one unaided item at the end; record what the unaided item shows and do not claim retention or fluency; give learners a job aid and items to practise later untested proposal
A fixed short time budget per session (e.g. 10 or 15 minutes a day) Per session: 1–2 minutes of unaided items on earlier procedures, 3–5 minutes of model or worked example on the new step only if it is new, the rest supported or mixed practice; one new step per session at most untested proposal; spacing across sessions from Distributed Practice Improves Retention [+M] (verbal recall, carried)
Learners have low confidence or low literacy, or work in a second language Let learners show a step by doing it before explaining it; keep written instructions short with pictures of each step; let early attempts be private (checked by software or one-to-one) before anything is shown to the group untested proposal
Errors are costly on the job (safety, compliance, harm to a client) or stakes are high More supported rounds per step; release only after several consecutive unaided successes on representative tasks; a stage of supervised unaided work before full independence; judge readiness only on unaided performance untested proposal
The procedure is a performance for a real audience (a demonstration, a recital, a service to a client) Practise the procedure to accuracy first in private, then rehearse the whole performance under the audience's conditions (time, setting, interruptions) at least twice before the real one untested proposal
Learners are second-language learners automatizing grammar or speech Practise the form in the same kind of task the learner will use it in (meaningful communication, comprehension or production), since gains may stay in the skill practised Practice of grammar rules improves only the practiced skill… [+W] and Automatization of L2 skills requires extensive practice in a consistent environment… [+W] (both from one review, reported second-hand)

Of the nineteen rows, ten cite a claim, and only the computer-based-practice row tests practice of a procedure as such; the others carry a neighbouring finding (cognitive-load studies, one-to-one tutoring of children, one algebra study, mathematics problem types, verbal recall, second-hand reviews) to a setting it did not test, and say so. The rest are untested proposals.

Observation, state and explanation

What can be observed is an attempt under stated conditions: the item, the aids and help available, the help taken (record its level), each step's correctness, where the attempt stalled, the time taken, and what the learner did with feedback on the next attempt. "Knows the procedure", "has automatized it" and "understands it" are inferred states. Accuracy with help shows what the learner can do with that help; speed on blocked items shows fluency on that block, not choice among procedures; accuracy today does not show retention.

Response Plausible explanations Observation that could change the interpretation
Accurate during practice, fails the first unaided or mixed item The support or the blocking made the decision for them; the item differs more than intended; anxiety or time pressure Give a matched unaided item without time pressure, then a mixed item; ask which procedure applies and why before they start.
Accurate but slow, with visible step-by-step checking Steps still effortful (early in learning); unsure of accuracy and double-checking; a step is half-remembered Time a short set of identical items across two sessions; ask for a confidence rating on each; check whether slowness sits at one step.
Fast and accurate on practised items, wrong on a changed item Procedure learned without its conditions or reasons; items were too alike; learner pattern-matches surface features Pose an item where the surface matches but the procedure does not apply, and ask for an explanation by drawing or pointing as well as words.
The same step is wrong every time A misconception the feedback did not address; the feedback was not seen or understood; the step was never modelled at its decision point Ask the learner to talk through that step on a fresh item; give an item where the wrong rule and the right rule predict different results.
Accurate at the end of the session, much weaker a week later Too little practice after first accuracy; practice all in one block; interference from a similar procedure learned since Re-test with new items at the delay, and record every practice and exposure in between; compare a procedure that was mixed with one that was not.

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.

Evidence and qualifications

No claim compares the whole arc with explanation alone, varies the amount of practice after first accuracy, or tests when to add the reasons behind a procedure. The claims below test single links.

  • In one experiment with 256 undergraduates, computer-based lesson versions that included practice produced significantly higher posttest scores than versions without practice [+M]: undergraduates in a computer-literacy course, randomly assigned within pretest blocks to six versions of one lesson (randomized); the four versions with practice beat the two without on the posttest (means above 17 against below 15). No effect size, the feedback during practice is not described, and the horizon is not reported. It supports including practice; it says nothing about how much.
  • Worked examples reduce unnecessary search for novices. [+M]: in algebra learning, learners who studied worked examples performed better than learners who spent the same time solving conventional problems; the authors' interpretation is reduced means–ends search. One entry, q3, read from the abstract, no effect size. The claim's Discussion says the benefit is strongest for novices and can become redundant once the schema is held.
  • Contingent scaffolding improves learning more than fixed or absent support. [+M]: fourth and fifth graders tutored one-to-one on long division (8 per condition) solved more problems under fully contingent support than under fixed, partly contingent or no support, at an immediate test and one month later; a meta-analysis of dynamic assessment ranked explicit strategy training above scaffolding and scaffolding above coaching; a review notes most evidence is one-to-one and contingency is scarce in whole-class teaching. It supports adapting help to each attempt; it does not show a classroom-scale effect.
  • Process goals lead to better skill acquisition for novices than outcome goals. [+M]: high-school students learning writing revision who were assigned process goals outperformed those assigned outcome goals, and the results suggested shifting to outcome goals after mastery can help; a theoretical chapter synthesises related work including dart-throwing. Sample size and effect size are not given. The second entry is a chapter coded q3, which reads high for a theoretical source.
  • Part-task practice reduces cognitive load for absolute novices during initial skill acquisition. [+M]: for complex material with high element interactivity, novices performed better when first given isolated parts before the integrated whole (one experiment, sample not stated, no effect size). It supports staging for novices on complex procedures; it does not say when to join the parts.
  • Interleaving Improves Discrimination [+M]: mathematics problems practised shuffled across types produced much better delayed test performance than the same problems blocked, with worse performance during practice; a meta-analysis finds an overall benefit that depends on how similar the categories are, largest when they are confusable. No effect size is printed on the entries. It supports mixing similar procedures once each is accurate; the page's timing of that mix is a proposal.
  • 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]: seventy seventh graders who compared alternative equation-solving methods side by side gained more procedural knowledge and flexibility than peers who studied the same methods one at a time, and no more conceptual knowledge (abstract only); a meta-analysis of case comparison reports d = .50. It supports comparison for flexible procedure use, and qualifies any expectation that comparison alone builds understanding.
  • Distributed Practice Improves Retention [+M]: in verbal recall tasks, spaced presentation beat massed presentation at equal study time (47.3% against 36.7% correct across 271 comparisons), and in fact learning the best review gap grew with the test delay. It is about verbal and factual material, not procedures; the page carries it to procedural practice as a proposal.

What limits or qualifies the model:

Learners, tasks and outcomes differ across these claims (undergraduates on a computer lesson, children with a tutor, high-school writers, seventh graders in algebra, adults recalling words), so do not rank them by effect label. None measures the fluency (speed with accuracy) of a procedure, and none follows a procedure's retention beyond a month.

The evidence by question

Further evidence, not yet read against this model

Claims that bear on procedural learning but are not part of the model above, each with what it is and how far it bears.

Objective and learner-valued goal

The designer's objective is usually accurate, and where it matters fast, unaided performance of the procedure on representative tasks, with stated aids, at a stated horizon, and sometimes adaptation to changed cases. Elicit separately what the learner wants from it: to pass a certification, to do a task at work without asking a colleague, to stop feeling slow, to understand why the procedure works. Agreement and divergence are both observations. A learner who values getting through quickly may want the steps without the reasons and resist step 6; one who values understanding may resist repetition after first accuracy. Say why each phase is there, and record how the learner's goal changed what they practised.

For example, a new care worker may want to complete a hoist transfer without calling a senior, while the designer's objective is passing the employer's moving-and-handling assessment. Practice on the assessment checklist serves the designer's objective; one supervised transfer with a real colleague in the worker's own unit, then one unaided under observation, tests the learner's. Record both, so that a pass on the checklist is not read as the capability the learner came for.

What would revise this model?

The expectation should weaken if studies comparing the arc (model, supported attempts, faded support, mixed and spaced practice) with an equal time of explanation or unsupported practice, on comparable learners and with unaided outcomes at a delay, show no advantage. If mixing procedures before each is accurate turns out no worse than mixing after, step 7's ordering should go. If adding reasons and conditions before the steps are fluent helps as much as adding them after, step 6 should move earlier. If spacing studies on procedures (rather than verbal recall) show a different shape, step 8's arc should follow them. If part-task practice of procedures delays integration more than it helps novices, step 2 should narrow. Do not protect the model by calling every failure "not enough practice" or "support removed too early" after the fact.

Accuracy with help, unaided accuracy, speed, delayed retention, choice among procedures, transfer to changed tasks and use in the setting the learner values are separate claims. The present evidence establishes no amount of practice to fluency, no spacing schedule for procedures and no criterion for when a procedure is learned.

Claims

  • Explicit Instruction — often supplies the visible model and sequence that procedural learning needs
  • Scaffolding — helps learners perform procedures they cannot yet execute independently
  • Deliberate Practice — explains how procedures improve through focused repetition with feedback
  • Modeling — makes the performer's decisions visible before the learner's first attempt
  • Worked Examples — a written model of a procedure studied before the first attempt
  • Guided Practice — the supported attempts that follow the model
  • Scaffolding and Fading — how support for a procedure is withdrawn
  • Spaced Learning — spreading practice across sessions

Examples

Illustrative

Procedural Information — Stepwise guidance is provided just in time as learners perform a routine or method.

Fading Scaffolding — Hints, prompts, or partial steps are gradually removed as learners become more capable of independent execution.

Modeled routine followed by coached repetition — The teacher demonstrates a procedure, learners practice with immediate correction, and then support is reduced over successive attempts.

Part-task practice — Isolated, repeated practice of a step that must become automatic, joined back into the whole task.

Connecting Concepts to Procedures — Instruction links each step of a procedure to the concept that justifies it.

Key Sources

Page metadata
Field Value
type principle
id procedural-learning
title Procedural Learning
description For a novice on a procedure, a visible model, supported attempts with feedback, support withdrawn on unaided success, then mixed practice spread over sessions with the steps tied to their reasons are expected to yield accurate, fluent, retained and flexible unaided performance; no claim tests the arc as a whole, only single links in it.
status review
canonical True
generated by claude/unspecified on 2026-10-05
trust unverified — no one has confirmed this page's content
sources 1 — anderson-1982

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