The teacher has just shown a class how to solve a new kind of problem. The worked example was clear. The steps were visible. Several students nodded. Then the first independent question appears, and the room changes.
One learner starts correctly but freezes at the second decision. Another copies the surface pattern from the example even though the new question has changed. A third knows what to do but waits for confirmation after every line. Nothing unusual has happened. The explanation has ended, but independence has not yet begun.
This space between seeing and doing is where guided practice belongs.
Guided practice is the deliberately supported phase in which learners begin performing the target knowledge or skill while the teacher still observes, questions, prompts, corrects and adjusts support. Its job is not to make every attempt look successful. Its job is to convert a newly modelled route into increasingly independent performance without allowing confusion, guessing or teacher dependence to become the learner’s main method.
This guide continues the How Teaching Works series after How Worked Examples Work in Teaching and How Questioning Works in Teaching. It owns the general teacher-side mechanism of guided practice. Stage-specific applications, such as the existing Primary 2 Mathematics guide on scaffolding and transfer, remain separate application pages. The newer Tutor Handbook articles on support dose, fading and independence testing remain narrower tutoring-control treatments and are linked here rather than reproduced.
The classroom cases below are original teaching examples. They are not records of actual students, experimental trials or official examination questions. External evidence is used to support specific principles, not to claim that this exact article has been independently tested as one complete programme.
A route through guided practice
Start with what guided practice is, then examine how responsibility moves, how support is selected, how the teacher checks and corrects, and how support is faded. The worked cases show the mechanism in Mathematics, English and Science. Later sections cover small-group tuition, failure modes, lesson design, parent support and evidence boundaries.
Guided practice is neither demonstration nor independence
During modelling, the teacher carries most of the cognitive responsibility. The learner watches a process being made visible: what to notice, which decision comes first, why a step is valid, how an expert checks the result. During independent practice, the learner is expected to select and execute much more of that process without the teacher carrying the decisions.
Guided practice sits between those states. The learner now has to act, but the teacher has not withdrawn completely. The teacher can still stop the process before an error becomes rehearsed repeatedly, ask what the learner is trying to do, reduce the task, point attention to a relevant feature, provide a partial cue, or briefly remodel a missing step.
The Australian Education Research Organisation’s current model of learning and teaching places guidance, scaffold removal, checking for understanding and additional instruction within its instruction phase, while its gradual-release phase emphasises guided and independent practice over time. AERO also notes that these practices are interdependent rather than a rigid one-way script. See Teaching for How Students Learn: Practice Guides and its 2026 model animation.
The Education Endowment Foundation makes a similar practical distinction in its 2026 seven-step planning material: modelling is followed by checking, guided practice, independent practice and reflection, while the amount of guided practice varies with learner need. See Modelling Independence: The Seven-Step Model.
These models should not be treated as a command to make every lesson look identical. Some familiar tasks need very little guidance. Some complex tasks need repeated movement between modelling, guided attempts and brief independent checks. The teacher’s job is to match support to the learning state, not to perform a ceremonial “I do, we do, you do” sequence regardless of evidence.
The purpose is successful thinking, not successful-looking work
A learner can complete a page correctly while the teacher quietly supplies the difficult decisions. The teacher points to the relevant sentence, names the formula, reminds the learner of the first step, asks a leading question at the second step and confirms every line. The final work is correct. The measurement of independence is not.
This distinction is central to the existing eduKate principle that a supported answer is not the same measurement as an independent answer. Guided practice is valuable precisely because support is allowed. But the support must be visible to the teacher’s own judgement. Otherwise the teacher may withdraw it too soon or keep it long after it is needed.
Record the difference between what the learner produced, what the teacher supplied and what happened after the support. “Solved with one prompt to choose the representation” tells a different story from “solved independently.” So does “needed the first step modelled, then completed the rest.” These are not weaker records. They are more accurate.
The aim is not to eliminate help as quickly as possible. The aim is to use enough help to create productive action, then reduce help when the learner can carry more of the work. Support is a temporary engineering structure, not a moral failure and not a permanent feature of competent performance.
Responsibility should move at the level of decisions
Teachers often think of release in terms of how much of a worksheet the learner completes alone. A more useful view is to track who is making the important decisions. Who identifies the relevant information? Who chooses the representation? Who selects the method? Who checks whether the answer is plausible? Who decides what to do after an error?
Two students can write the same number of lines while carrying very different amounts of responsibility. One learner may generate every decision and simply receive a final accuracy check. Another may wait for the teacher to nominate each next move. The written product hides the difference.
A practical release sequence can therefore be described by decision ownership:
- Teacher-owned: the teacher models the decision and explains why it is made.
- Shared: the teacher asks the learner to choose between plausible next moves and justify the choice.
- Learner-led with prompt: the learner proposes the next move; the teacher supplies a cue only if the process stalls.
- Learner-owned with delayed check: the learner completes the decision chain before receiving feedback.
- Learner-owned across variation: the learner chooses the method when surface features change and no cue announces which strategy applies.
Release does not have to happen evenly across every component. A learner may independently choose the correct method but still need arithmetic support. Another may calculate accurately but need help deciding what the problem is asking. Keep the learning target stable while changing support around the part that is not yet secure.
This is one reason diagnostic teaching and guided practice are inseparable. The teacher needs evidence about the learner’s present state in order to decide what responsibility can be transferred next.
Use the smallest support that reopens useful thinking
When a learner is stuck, the easiest response is often to explain the whole step again. Sometimes that is correct. Sometimes it replaces thinking the learner could have performed with a smaller cue. A useful support ladder starts with the least intrusive intervention that can reasonably restart the process.
Pause. The learner may simply need time. Do not convert a three-second silence into a teacher-led solution.
Ask for the current state. “What have you established so far?” or “Which part are you sure about?” This can reveal whether the learner is actually stuck or merely uncertain.
Direct attention. Point to the condition, diagram feature, sentence or quantity that should be considered. This does not yet supply the operation.
Offer a choice. “Would you compare the values directly or rewrite them in a common form?” A choice narrows the search while still requiring selection and justification.
Provide a cue. Remind the learner of a principle without applying it fully: “What has to happen to both sides of an equation?”
Model the missing part. When the knowledge is absent or the learner is repeatedly guessing, show the step clearly. Then return responsibility on the next part or a fresh task.
EEF’s 2026 scaffolding guidance describes a related principle of providing the least amount of support first and increasing support only when needed. See How to use visual, verbal and written scaffolding to support pupils’ use of metacognitive strategies. The practical implication is not that teachers must follow one universal prompt hierarchy, but that scaffolding should remain responsive and temporary.
A smaller prompt is not automatically better. If the learner lacks the prerequisite knowledge, a sequence of increasingly obvious questions can become a guessing game. At that point direct teaching is more efficient and more respectful. The rule is not “never tell.” The rule is “do not hide teaching inside a chain of prompts and then misread the result as independent performance.”
Choose practice items for the decision you want learners to rehearse
The first guided task should usually preserve the important structure of the model while changing enough surface detail to require a new act of thinking. If the task is identical except for one number, it may be useful for procedural rehearsal. If the next task changes every feature at once, failure may tell the teacher little about which change caused difficulty.
Early examples can be deliberately close. Later examples should vary the conditions that define the concept. A learner who can expand only expressions that look exactly like the teacher’s example has learned less than a learner who can identify the distributive structure across different coefficients, signs and arrangements.
AERO’s current Vary Practice guide emphasises spacing and varying opportunities so learners can retain and apply knowledge across contexts. Guided practice should therefore be designed as a trajectory: initially stabilise the new route, then introduce variation that requires recognition and selection.
Do not confuse variation with random difficulty. The teacher should know what changed and why. If the lesson is testing method selection, vary the cues that normally announce the method. If it is testing interpretation, vary the wording while preserving the conceptual structure. If it is testing transfer, change the context after the original relationship is secure enough to travel.
Check early enough to prevent long rehearsals of the wrong route
Guided practice is not independent work with the teacher standing nearby. The teacher samples the process while there is still time to correct it. Waiting until the end of twenty near-identical questions can turn one misunderstanding into twenty rehearsals.
The check should target a consequential decision. Ask learners to show the first line, name the method, identify the evidence, draw the representation or explain the rule being used. The teacher is not looking for constant interruption. The aim is to inspect the points where a wrong decision would contaminate the rest of the task.
AERO’s Monitor Progress guide, updated in May 2026, frames checking for understanding as a way to identify what students can do and provide additional instruction, guidance or feedback where necessary. That principle is central to guided practice: evidence is gathered because the next teaching move is still adjustable.
Corrective feedback should be specific enough to change the next attempt. “Careless” does not tell the learner what relation was missed. “Check the sign before you distribute the negative coefficient” points to a feature that can be inspected. “Use more evidence” is weaker than “Your claim concerns motive; which action in the passage supports it?”
After correction, give the learner work to do. Ask for the repaired step, a new example or an explanation of why the correction matters. A correction that exists only in the teacher’s voice has not yet become part of the learner’s performance.
Do not rescue every error before it becomes visible
There is also a danger in correcting too quickly. If the teacher interrupts at the first sign of hesitation, the learner never has to monitor, detect or repair. Guided practice should leave space for productive self-correction when the learner has enough knowledge to perform it.
A practical distinction is between an error that will teach the teacher something useful and an error that is simply being repeated. Let the learner commit to a method long enough for the reasoning to become inspectable. Then ask for a check. If the learner can detect the problem, the self-monitoring process deserves reinforcement. If not, intervene before repeated practice stabilises the wrong procedure.
This is not a fixed timing rule. A hazardous practical procedure, a foundational misconception or an error likely to spread across many subsequent steps may justify immediate interruption. A reversible calculation slip may be better left for the learner to catch through substitution or estimation.
The broader principle is to design opportunities for the learner to use the checking method that independent performance will later require. A learner who succeeds only because the teacher catches every mistake has not yet learned to regulate the process alone.
Fade support by removing information, not by simply becoming quieter
A teacher can speak less while still carrying the important decisions through gestures, sequencing or task design. True fading reduces the amount of problem-solving information supplied by the support. The learner must increasingly decide what to notice, what to retrieve and what to do next.
Remove one support at a time when practical. If the worked example contained the full method, the first guided task might retain a partially completed structure. The next can remove the intermediate steps. Later the method label disappears. Eventually the learner meets a mixed set in which choosing the method is part of the task.
Fading should be tied to evidence rather than the clock. “We have practised this for ten minutes” is not evidence that the support can be removed. Look for accurate execution, explanation of the critical relationship, reduced need for prompting and success on a fresh item where the answer is not cued by the previous example.
Some learners will need different durations of guided practice. EEF’s 2026 seven-step model explicitly notes that pupils may require varying amounts before independent practice. That variation should not be converted into fixed ability labels. The amount of support needed today is a state, not an identity.
When support is removed and performance collapses, do not treat that as proof the original teaching failed completely. The collapse reveals what the scaffold had been carrying. Reintroduce a smaller amount of support, target the missing component and attempt release again. The Tutor Handbook’s The Fade develops this tutoring control problem in more detail.
Worked case: solving an equation without turning the teacher into the first step
Suppose learners have just seen a worked example for an equation such as 3(x + 4) = 24. The teacher modelled two valid routes: divide both sides by 3 first, giving x + 4 = 8, or expand first, giving 3x + 12 = 24. The teaching goal is not merely to obtain x = 4. It is to preserve equality while choosing and executing a valid sequence.
For the first guided item, use 5(y + 2) = 35. Ask learners to write only the first transformation. Do not name the route. If a learner writes y + 2 = 7, ask what operation was applied to both sides. If a learner writes 5y + 2 = 35, the issue is not the final arithmetic; the distribution has been applied incompletely.
At this point the teacher can use the smallest useful support. “What is multiplying the whole bracket?” directs attention without supplying the corrected line. If the learner still cannot proceed, briefly remodel distribution with a simpler expression and return to a fresh equation.
The next item can remove one support: 4(a – 3) = 20. Now the sign inside the bracket changes. A learner who divides first obtains a – 3 = 5, then a = 8. A learner who expands obtains 4a – 12 = 20, then 4a = 32 and a = 8. Both routes are valid. The teacher can ask which route felt more efficient and why without insisting on one universal method.
Then change the structure: 3(2m + 1) = 21. Dividing by 3 gives 2m + 1 = 7, then 2m = 6 and m = 3. The learner now has to manage an additional coefficient inside the bracket. If the teacher announces “divide by 3 first,” method selection is still supported. If independence is the target, the method cue should eventually disappear.
A final guided-to-independent transition can mix equations that invite different convenient routes. The question becomes not only “Can you solve?” but “Can you choose, execute and check?” Substituting the solution into the original equation provides a learner-owned check. Guided practice has now moved from imitation towards controlled decision-making.
Worked case: building an evidence-based inference one decision at a time
Use this original passage: Nadia placed the unopened envelope beneath her notebook before the others entered. During the meeting she answered every question about the project, but when the chairperson mentioned the competition results, she looked at the clock and began packing her bag.
The modelled task has already shown learners how to separate a textual detail from an inference and how to avoid claiming more than the passage supports. During guided practice, the teacher should now transfer those decisions in stages.
First ask each learner to underline two details that could matter. If a student underlines the envelope and Nadia’s reaction to the results, the teacher can ask what relationship might connect them. If the learner selects only generic details such as “she answered every question,” ask whether that detail helps explain her response to the competition results.
A learner may infer that Nadia is worried about the result contained in the envelope, but the text does not explicitly say what the envelope contains. The teacher can guide precision by asking, “Which part is observation and which part is your interpretation?” The support is focused on claim strength rather than supplying a preferred answer.
Next, remove the evidence-selection prompt. Give a fresh short passage and ask learners to write a claim, one supporting detail and one sentence explaining the link. The teacher can still circulate and ask questions, but should resist naming the evidence before the learner has attempted selection.
Finally, the learner meets an unfamiliar passage in which no label announces that inference is being tested. The decision to distinguish statement from implication is now part of the performance. This is the point at which guided practice begins to become transfer.
Worked case: designing a fair comparison without supplying every control
Imagine a science lesson in which learners are designing a simple investigation to compare how quickly two materials allow water to pass through. The teaching goal is to reason about fair comparison: change the intended factor, keep relevant conditions comparable, measure an outcome and avoid claiming more than the setup can show.
The teacher has modelled one example in a different context. In guided practice, learners now receive two filter materials, equal containers and the task of comparing the amount of water collected after a fixed time. Rather than list every control, the teacher asks: “What must be the same if the material is the factor we want to compare?”
One learner identifies the starting volume of water. Another names the area of material exposed. A third proposes using different pouring times because one material appears slower. The teacher asks what changing the pouring time would do to the comparison. The group now has to connect a design choice to interpretability.
If the group misses an important condition, the teacher can direct attention: “Look at the two containers. Is there anything about the path the water travels that differs?” This cue preserves more learner responsibility than immediately naming the difference. If learners still cannot identify it, direct teaching is appropriate.
The next task changes context. Learners compare cooling in two containers and must decide what should be kept comparable. The surface has changed, but the reasoning structure is related. Guided practice now checks whether the concept of fair comparison can travel beyond the original apparatus.
Do not treat these examples as a universal experimental template. Real investigations differ in what variables matter, what can be controlled and what claims are justified. The teaching point is the transfer of a reasoning relation, not a ritual list of “controls.”
Guided practice in a three-student group
A small group makes learner thinking more visible, but it also makes over-support easy. The teacher can give so much moment-to-moment help that all three students appear successful while each is carrying a different fraction of the actual decision-making.
Use a shared intellectual object but separate first attempts. Each learner writes the first step, selects evidence or states the prediction before hearing the others. The teacher can then compare the responses and decide whether the next move should be common or individual.
One learner may need a prerequisite briefly retaught. Another may need only a prompt to check a condition. A third may be ready for a variation. This does not require three unrelated lessons. The group can return to a common problem after each learner receives the smallest support needed to participate productively.
Rotate who explains first. Otherwise the fastest learner can become the permanent model and the other two receive peer-supplied scaffolding before they have attempted the decisions themselves. Shared discussion is valuable, but it should be followed by opportunities to see what each learner can now do.
This connects with How Can Three Students Share a Class Without Receiving the Same Teaching? and How Group Study Works. The central principle is the same: participation is shared, but learning must eventually survive at the individual level.
Guided practice can include collaboration without outsourcing thinking
Pair or group work can be useful during guided practice when the task requires learners to compare reasoning, explain a decision or inspect alternatives. But the group should have an intellectual job beyond producing one answer.
For example, ask one learner to propose a method, another to identify the condition that makes it valid, and a third to test the result. Then rotate roles. The structure forces multiple kinds of reasoning into the discussion rather than allowing one student to solve while others copy.
After collaboration, return to a fresh individual item when individual competence is the claim being made. A group answer is evidence about what the group produced under shared conditions. It is not automatically evidence that every member can reproduce the reasoning alone.
The teacher should also listen for misconceptions moving through the group. Peer explanation does not become accurate merely because it is peer-generated. Guided practice still requires a knowledgeable teacher to protect the disciplinary boundary.
Different scaffolds carry different parts of the task
A scaffold can reduce memory demand, narrow a decision, organise information or provide a partial representation. Teachers should know what each scaffold is carrying, because that is what the learner will eventually need to carry after the scaffold is removed.
A worked step carries part of the procedure. A checklist carries sequence and monitoring. A sentence frame carries part of language structure. A diagram carries a representation. A prompt card carries reminders about decisions. A partially completed table carries organisation.
The scaffold should match the barrier. A learner who understands the science but struggles to organise a comparative answer may benefit from a structure cue. A learner who lacks the scientific relationship itself needs teaching, not merely a better writing frame. A mathematics learner who forgets a multi-step sequence may benefit from a checklist, while a learner who chooses the wrong method needs work on recognition and selection.
Accessibility support should not automatically be treated as cognitive support. A learner may independently reason while using a visual access tool, enlarged text, an agreed response format or other appropriate support. The question is which part of the target performance the aid supplies. This distinction keeps independence claims fair and precise.
Guided practice should make metacognitive decisions visible
Experts do more than execute procedures. They monitor whether the approach is working, notice contradictions, decide when to switch methods and check whether an answer fits the original problem. Novices often need these decisions externalised before they can regulate them independently.
During guided practice, the teacher can ask: “What are you checking here?” “What would tell you this method is failing?” “Which part of your answer is least certain?” “What could you do before asking me?” These questions do not replace subject knowledge; they help the learner manage its use.
EEF’s current 16–19 metacognition guidance explicitly recommends modelling and guided practice of metacognitive strategies before independent use. See Metacognition and Self-Regulation. The important boundary is that metacognitive prompts should stay embedded in real subject tasks rather than becoming detached slogans about “thinking about thinking.”
The learner-side system is developed in How Metacognition Works in Learning. Guided practice is where the teacher can initially carry part of that monitoring process and then hand it over.
Common failure modes in guided practice
The teacher solves through questions. Every prompt contains enough information that the learner only has to agree. The final answer looks student-generated, but the decision path was teacher-owned.
Independent practice begins too early. Learners are sent to a worksheet after one modelled example even though they have not yet made the important decisions with support.
Guided practice never ends. The teacher confirms every line and never creates a clean attempt without support. Learners become accurate in the teacher’s presence but fragile when the prompt disappears.
The scaffold carries the target skill. A sentence frame supplies the reasoning the learner is supposedly practising, or a worksheet names the method when method selection is the learning goal.
Errors are corrected but not re-attempted. The learner copies the teacher’s repair and moves on. There is no fresh evidence that the relationship has become usable.
Practice is too uniform. Learners become fluent at a template but cannot recognise the same structure when wording, representation or context changes.
Time is mistaken for readiness. Support is removed because the lesson plan says the class should now be independent, not because evidence shows they can carry the decisions.
Support becomes a label. A learner who needed heavy guidance yesterday is permanently treated as someone who always needs it. Guided practice should update with performance, not preserve old classifications.
Design a lesson around changing responsibility
A useful lesson plan can specify not only tasks but responsibility states. For each important step, ask: what will the teacher demonstrate, what will the class perform together, what will learners attempt with prompts available, and what will they later perform without those prompts?
Begin by retrieving the prerequisites the new task depends on. If those foundations are missing, guided practice on the advanced skill will be unstable. Teach or repair the prerequisite first.
Model the new decision clearly. Then choose one or two early practice items that are close enough to the model to focus attention on the new relationship. Collect visible responses before learners complete large amounts of work.
Plan likely errors. Decide which errors call for a cue, which call for a brief remodel, and which can be left for a learner-owned check. This turns guidance into a designed response system rather than a series of improvised hints.
Plan the fade. Remove a step, a prompt, a method label or immediate confirmation. Then include a fresh item to test whether the learner can carry the missing part. If performance drops, diagnose what the removed support had been doing.
End with evidence that can inform the next lesson. An independent exit item can be useful, but it should match the learning goal. If the target is method selection, the item must not announce the method. If the target is explanation, a correct numerical answer alone is insufficient.
Guided practice across several lessons
Release is not only a within-lesson event. A learner may perform independently at the end of one lesson while the structure is still fresh, then need renewed support when the idea returns days later. That does not invalidate the earlier learning; it shows that accessibility across time is another dimension of independence.
Revisit the knowledge after delay. Start with a short retrieval or application before displaying the old scaffold. If learners can reconstruct the route, keep the support absent. If they can recover with one cue, use the cue and try again later. If the whole structure has disappeared, more substantial reteaching may be needed.
Over several lessons, increase variation and delay while reducing cues. The path is not simply “more questions.” It is a change in what the learner must notice and decide without external help.
This trajectory connects guided practice to retrieval, spacing and transfer in How Teaching Works and to the cognitive-load boundary in How Cognitive Load Works in Learning.
When guided practice should become independent practice
There is no single percentage that defines readiness. The decision should reflect the importance of the next task, the learner’s accuracy, the quality of reasoning and the amount of support still being used.
Useful signs include accurate execution on more than one item, explanation of the critical relationship, successful checking, reduced reliance on prompts and performance on a fresh task where the method is not announced. None needs to be perfect. The teacher is deciding whether an unsupported attempt is now likely to be productive and informative.
Independent practice is itself part of learning, not merely a test after teaching. Learners need opportunities to consolidate, retrieve and apply without the teacher carrying decisions. If every independent attempt is withheld until perfection is guaranteed, independence never gets practised.
When independent work reveals a gap, return to guided practice selectively. Do not automatically restart the whole sequence. Identify the component that failed and restore only the support needed to rebuild it.
For teachers: a compact guided-practice planning check
- What exactly should the learner be able to do after the model?
- Which decisions are still likely to need support?
- What will the first learner action be?
- How will I see that action early?
- Which prompts preserve thinking, and which prompts would supply the answer?
- What error would require immediate reteaching?
- What scaffold am I using, and what part of the task does it carry?
- What will I remove first?
- What fresh task will show whether the removed support is no longer needed?
- How will I distinguish supported success from independent success?
The list is not a compliance form. Its purpose is to force the key instructional decisions into view before the lesson becomes a stream of spontaneous hints.
For parents: help without becoming the method
Homework often creates an accidental guided-practice environment. A child is stuck, the parent wants to help, and the easiest path is to lead the child through every step. The work gets finished, but the parent becomes part of the procedure.
Start by asking what the child can already do. “Show me where you got to without help” preserves more information than beginning with a full explanation. Then use a small prompt that targets the exact barrier: reread the condition, identify the relevant example, explain what a symbol means or choose between two possible first steps.
If the child lacks the knowledge, explain it clearly. Do not turn the evening into a twenty-question guessing game. After helping, give a small fresh attempt if appropriate. The purpose is to discover what the child can now carry, not simply to finish the original page.
For the wider relationship between home support and independence, see How Homework Works in Learning.
Using AI during guided practice
AI can generate examples, hints, explanations and checks quickly. That makes it useful as a practice tool and dangerous as an invisible decision supplier. If the system always names the next step, method or evidence, the learner may produce stronger work without learning to make those decisions.
Use AI support in levels. Ask first for a question only. If needed, request a general cue rather than the next step. Escalate to a more explicit hint when the learner is genuinely blocked. After a worked explanation is shown, close it and attempt a fresh problem without the same support.
Teachers should verify generated content before using it. A fluent hint can contain an incorrect assumption or accidentally reveal too much. The same support-dose logic applies whether the scaffold comes from a person, a worksheet or a model.
The broader learner-side boundary is developed in How AI-Assisted Study Works: help should leave the learner stronger, not merely improve the current output.
A guided-practice record should describe state, support and next test
A useful teaching note can be short: “Selected the correct method after attention cue; executed accurately; next lesson test method selection without cue.” Another might read: “Understood inference/evidence distinction after model; still needs prompt to limit claim strength; use fresh passage.”
These notes avoid two errors. First, they do not exaggerate supported performance into mastery. Second, they do not turn temporary support into a permanent identity. The note points to a next test that can update the learner model.
Keep identifiable records proportionate to educational need and within appropriate authorised systems. The purpose is instructional continuity, not the accumulation of unnecessary personal speculation.
Evidence, interpretation and limits
This guide combines external educational guidance with original examples and an eduKate teaching framework. The external sources support specific principles: teacher modelling followed by guided and independent practice, checking understanding, adjusting support, scaffolding with a plan to remove it, and varying practice over time. They do not establish that every sequence, prompt or example in this article is optimal for every learner or subject.
AERO’s current model and practice guides are useful because they connect explicit instruction, guidance, progress monitoring, practice and gradual release. EEF’s current materials provide practical scaffolding and metacognitive planning guidance. Older IES materials also describe guided and scaffolded practice within explicit instruction, especially in mathematics intervention contexts. See the IES instructional-process resource for a historical example of this framing.
Evidence from a particular intervention, age group or subject should not be generalised without qualification. A technique that helps novice learners acquire a structured procedure may require adaptation for advanced inquiry, creative composition, practical performance or learners with different prior knowledge. The level of guidance should follow the actual task and learner state.
Sources for this edition were reviewed on 12 September 2026. Currentness matters because organisations update guidance and examples over time. The date of a webpage update should not be confused with the date of every underlying study it cites.
Selected sources
Australian Education Research Organisation: Teaching for How Students Learn: Practice Guides; Monitor Progress, updated 14 May 2026; Vary Practice, updated 14 May 2026; and the 2026 model animation.
Education Endowment Foundation: Modelling Independence: The Seven-Step Model, 2026; How to use visual, verbal and written scaffolding to support pupils’ use of metacognitive strategies, 2026; and Metacognition and Self-Regulation.
Institute of Education Sciences: The Instructional Process in Interventions, an older explicit-instruction resource describing modelling, guided/scaffolded practice, feedback and movement towards independence.
Guided practice ends when the learner can carry what the support used to carry
Return to the learner who froze on the first independent question. The answer is not automatically “more explanation” and not automatically “leave them to struggle.” The teacher needs to discover which decision failed, provide enough structure for useful action, and then test whether that structure can be reduced.
That is the central mechanism. Model the route. Let the learner act. Observe the first consequential decision. Support only what needs support. Correct before error becomes routine. Give the learner another attempt. Remove part of the scaffold. Change the surface. Delay confirmation. Return later. Watch for the moment when the teacher’s information is no longer required.
Guided practice succeeds when support creates capability that survives the removal of support.
Continue through How Teaching Works, revisit How Worked Examples Work in Teaching and How Questioning Works in Teaching, or move into the tutoring-control layer through The Dose, The Fade and The Independence Test.
