A teacher has planned the lesson carefully. The explanation is clear, the examples are sensible, and the practice sequence is ready. Ten minutes in, the evidence says the plan is no longer the best next move.
Some learners have understood the new relation and are ready to apply it. Others are using the wrong rule. One learner can perform the method only while the example remains visible. Another already knows the content and is becoming less attentive because the practice is redundant.
The teacher now has an adaptive teaching problem.
Adaptive teaching works when teachers gather useful evidence about current learning, interpret what that evidence means, and change explanation, support, practice, grouping, representation, pacing or extension in ways that improve access to the same worthwhile learning goal rather than quietly lowering the goal for whoever is struggling.
This article continues the eduKate Sengkang How Teaching Works series after How Transfer Works in Teaching. It owns the teacher-side architecture of evidence-driven adaptation.
It does not replace subject-specific differentiation guides, the Tutor Handbook’s adaptation and regrouping gates, or learner-side metacognition pages. Those routes retain their narrower jobs.
The central question here is simple: what should a teacher change when the learners’ evidence changes?
A route through adaptive teaching
Begin with what adaptive teaching is, then move through evidence, interpretation, decision types, keeping the goal intact, support adaptation, grouping, extension, accessibility, and verification. Worked cases cover Mathematics, Science and English, followed by whole-class, small-group, homework, examinations, AI and failure modes.
Adaptive teaching is evidence-driven variation in the next move
Adaptation begins with evidence, not preference.
The teacher notices what learners know, misunderstand, retrieve, fail to retrieve, can perform with support, can perform independently, or are ready to extend. The next teaching move changes because the current state differs from the planned assumption.
EEF’s 2026 Check. Adapt. guidance frames this around three broad pathways: many misunderstand, so pause and fix; some are unsure, so adapt support; most understand, so extend and support. The guidance explicitly presents these as decision supports rather than a rigid formula.
AERO’s Monitor Progress guide similarly emphasises checking what learners understand and can apply, then providing additional instruction, guidance or feedback where needed.
Adaptive teaching is not improvisation without a learning model
Changing activities constantly is not the same as adapting well.
The teacher needs a model of the intended learning, likely prerequisites, common errors, useful evidence and plausible response options. Good preparation makes adaptation easier because the teacher knows what can be changed without losing the objective.
Adaptation is disciplined flexibility.
Adaptive teaching is not a separate curriculum for every learner
Most classrooms and tuition groups need shared intellectual objects: common texts, concepts, problems, examples or goals.
Adaptation changes the route, support, representation, amount of practice or level of extension while preserving coherence.
Three learners do not need three disconnected curricula simply because they need different next moves.
Adaptive teaching is not “learning styles”
Teachers should not assign fixed learner identities such as visual, auditory or kinesthetic and then adapt everything to those labels.
Representations should be chosen because they fit the content and the current learning barrier, not because a learner is assumed to belong to a permanent cognitive type.
The learner model should remain revisable.
Adaptive teaching is not automatically easier teaching
A learner who is struggling may need clearer explanation, a restored prerequisite, a better representation or more guided practice.
That does not automatically mean a lower intellectual goal.
Sometimes the adaptation should remove unnecessary difficulty so the learner can engage with the real challenge.
Adaptation depends on evidence that can change a decision
Useful evidence can come from questions, mini-whiteboards, retrieval attempts, written work, worked examples, explanations, error patterns, homework, observations, practice papers and fresh independent tasks.
The teacher should know what each source can and cannot show.
A fluent class discussion may hide individual gaps. A low score may combine knowledge, timing and reading problems. A successful supported answer does not prove independence.
Broad response sampling matters
If adaptation is based only on volunteers, the teacher adapts to the most visible learners.
Use response methods that make enough of the class visible when a consequential decision is being made.
Mini-whiteboards, short individual answers, simultaneous cards, digital polls and silent first responses can all help depending on the task.
Evidence should preserve support conditions
Record whether the learner answered independently, after a category cue, after a peer discussion, with the worked example visible, or after teacher modelling.
The same correct answer means different things under different support conditions.
One answer rarely justifies a large route change
Use proportionate decisions.
A one-off slip may need a correction. A repeated misconception across related tasks may justify reteaching. A pattern across several lessons may justify a larger change in route or grouping.
Interpret the first breakdown, not only the final error
A wrong final answer can begin with task interpretation, missing knowledge, method selection, execution, checking or transfer.
Adaptive teaching improves when the teacher identifies the earliest consequential breakdown.
Changing the wrong part of the lesson can add effort without repairing the actual barrier.
Separate missing knowledge from wrong knowledge
Missing knowledge calls for teaching. Misconception calls for model repair. Weak retrieval calls for retrieval and spacing. Prompt dependence calls for fading. Transfer failure calls for variation and selection practice.
The next move should match the state.
Separate access barriers from target-skill barriers
A learner may understand the concept but struggle with dense language, small print, unfamiliar vocabulary or a response format.
Adapt access without quietly changing the intellectual target.
Separate pace from understanding
Fast answers are not automatically deep. Slow answers are not automatically weak.
Adapt time only when pace is part of the barrier or future performance condition.
Separate confidence from competence
A confident wrong answer may require misconception repair. A correct uncertain answer may require retrieval and confirmation. A quiet learner may understand well.
Do not adapt to confidence alone.
Adaptive teaching has a finite set of common decision types
- Pause and fix: stop progression because a prerequisite or misconception is widespread.
- Clarify: rephrase or represent the same idea more clearly.
- Model: reveal a hidden process or decision.
- Scaffold: add temporary support while preserving the target.
- Fade: reduce support because capability has grown.
- Regroup: temporarily change who works together because needs differ.
- Change practice: increase, reduce or vary examples.
- Restore prerequisite: revisit earlier knowledge that the current lesson depends on.
- Extend: deepen, vary or transfer when learners are ready.
- Change evidence: use a better question or task because current evidence is ambiguous.
- Delay decision: gather more evidence before making a large route change.
Pause and fix when the next step depends on the unresolved idea
If most learners misunderstand a foundational relation, moving on can multiply the error.
Reteach, contrast examples, check again, then resume when enough of the class can participate meaningfully.
Adapt support when some learners are unsure
Do not force the entire class back to maximum support because a subset needs it.
Use optional worked examples, cue cards, teacher conferencing, targeted mini-groups or different completion steps.
Extend when most learners are secure
Extension should deepen the same worthwhile learning, not merely give more volume.
Change representation, add a constraint, require comparison, ask for another method, use transfer, or ask learners to justify a boundary.
Keep the core learning goal intact unless the evidence justifies changing it
Adaptation can become quiet curriculum narrowing if struggling learners repeatedly receive simpler goals.
Ask whether the goal itself is inappropriate or whether the route into the goal needs changing.
Preserve ambitious learning where possible, while adjusting prerequisites, examples, support or time.
Adapt the route before reducing the destination
Use clearer examples, smaller steps, better representations, guided practice or additional time before deciding that a learner needs a permanently lower target.
Some goals legitimately differ
Not every learner is always working towards the same immediate target. Different courses, accommodations, prerequisite states or individual plans can justify different goals.
The principle is transparency: know when the goal changes and why. Do not hide a goal change inside a scaffold.
Adapt support by function
Ask what support needs to carry: representation, retrieval, sequence, language, checking, planning or confidence.
Then choose a support that targets that function and preserves the rest of the learner’s work.
Change representation when the current form hides the relationship
Move between concrete, diagrammatic, verbal, tabular, graphical and symbolic forms where appropriate.
The goal is not to match a learning style. It is to expose the relevant structure.
Change example sequence when the boundary is unclear
Add a contrasting case, a non-example, a simpler case or a case where the tempting wrong rule fails.
Do not simply add five more examples of the same surface pattern.
Change question type when the evidence is weak
If multiple-choice answers do not reveal reasoning, ask for a justification. If long written answers add language noise, use a short diagram or oral explanation first.
Change support dose rather than changing the whole task
One learner may need a category cue; another may need the first step modelled; another may need no support.
See How Scaffolding and Fading Work in Teaching.
Change feedback timing
Novices repeating a foundational error may need immediate correction. More experienced learners may benefit from time to self-check before feedback.
Change the amount of practice
Some learners need more repetitions; others need variation rather than volume.
Once a procedure is fluent, endless similar questions can waste time.
Change the spacing interval
Fragile learning may need an earlier return. Stable learning can move into longer maintenance intervals.
See How Spacing Works in Teaching.
Grouping is a temporary instructional decision, not a learner identity
Group learners because a shared next move is useful, then regroup when the evidence changes.
Do not let yesterday’s performance become permanent grouping destiny.
Regroup around current barriers
One temporary group may revisit a prerequisite. Another may practise transfer. Another may compare multiple methods.
All can return to a common class task afterward.
Avoid fixed “high, middle, low” roles inside small groups
A learner can be strong in one topic and need support in another.
Role and grouping should follow the task state.
Peer support needs individual accountability
Collaboration can help explanation and repair, but group output does not prove individual understanding.
EEF’s August 2026 independent-learning guidance on collaborative approaches stresses deliberate design, participation and individual accountability.
Extension should deepen, not merely accelerate
Faster learners do not always need the next year’s content immediately.
Ask for another method, a proof, a counterexample, a stronger explanation, a changed representation, a transfer case or a boundary condition.
Depth can preserve curriculum coherence while maintaining challenge.
Do not manufacture weaknesses in strong work
If a learner has met the goal well, adapt by extending the intellectual problem rather than inventing arbitrary corrections.
Accessibility adaptations should preserve the construct
Text-to-speech, enlarged text, extra processing time, alternative response modes and assistive technology may be legitimate access supports.
Ask whether the support gives access to the task or supplies the target reasoning.
Adaptation should remove avoidable barriers without erasing what the learning goal is supposed to measure.
Language adaptation should distinguish concept from expression
A multilingual learner may understand a Science concept but need vocabulary support to express it in English.
Use diagrams, glossaries or first-language reasoning diagnostically where appropriate, then return to the required English performance if English expression is part of the final goal.
Behaviour can be evidence about task fit without becoming a diagnosis
Off-task behaviour may reflect boredom, overload, misunderstanding, social dynamics, fatigue or unrelated causes.
Adjust only after gathering enough evidence to justify a learning-related interpretation.
Worked case: Mathematics adaptation after mixed responses
The class is solving 3(x + 2) = 21.
Some learners divide by 3 first and obtain x + 2 = 7, then x = 5. Others expand correctly to 3x + 6 = 21. Several write 3x + 2 = 21.
The error pattern suggests an unstable distributive relationship rather than a general equation problem.
Adaptation: pause the affected learners, compare 3(x + 2) with 3x + 2 using substitution at x = 4. The first equals 18; the second equals 14. Rebuild distribution, then use a fresh bracketed expression.
Learners who were already secure can compare the divide-first and expand-first methods and discuss efficiency.
The goal—solving equations while preserving equality—remains common. The route differs.
Worked case: Science adaptation when data reading and mechanism diverge
A class interprets a temperature-time graph. Some learners read the values accurately but overclaim that insulation stops energy transfer. Others misread the graph scale before reaching the mechanism.
Do not give both groups the same reteaching.
Group A needs a mechanism contrast: slower transfer versus zero transfer. Group B needs graph-scale interpretation. Then both return to the same fresh thermal-energy problem.
Worked case: English adaptation when inference and expression diverge
Three learners answer an inference question.
One selects relevant evidence but cannot explain the link. One writes fluent prose built on unsupported inference. One gives a strong answer quickly.
Adaptation: the first receives an evidence-to-claim prompt, the second compares supported and unsupported interpretations, the third receives a passage with deliberately ambiguous evidence and must calibrate certainty.
All continue working on evidence-constrained inference.
Adaptive teaching in a three-student group
Three learners make adaptation practical because the teacher can preserve a shared task while varying support almost immediately.
Collect individual first attempts. Identify the first unstable operation for each learner. Give the smallest distinct next move. Rejoin the group around a common comparison or transfer problem.
This prevents “small group” from becoming “same teaching delivered to fewer people.”
Adaptive teaching at whole-class scale needs preplanned options
Teachers cannot design every alternative route live.
Preplan one or two likely misconceptions, one support option, one extension option and one hinge question. The exact path remains evidence-driven.
Homework can reveal whether the in-class adaptation was enough
When teacher support disappears, the learner may expose a hidden dependence.
Use homework evidence to update the next lesson rather than treating homework errors only as compliance problems.
Examination preparation requires adaptation by error class
Do not respond to every low practice-paper score with “do more papers.”
Classify errors: missing knowledge, weak retrieval, task interpretation, method selection, execution, time, checking or transfer.
Adapt the training block to the dominant barrier.
AI can help generate adaptive options but should not make opaque learner diagnoses
AI can create simpler examples, alternative representations, practice variants and extension tasks.
Teachers should verify content and avoid treating a handful of responses as proof of a fixed learner type, intelligence level or psychological condition.
Use AI as a content and option generator; keep instructional judgement accountable to observable evidence.
Adaptation should be reversible
A good adaptation includes a future check.
If extra scaffolding was added, test whether it can fade. If the interval was shortened, see whether later retrieval stabilises. If the group changed, see whether the learner can rejoin common work.
Adaptation should have an exit condition
Temporary support should not become permanent by inertia.
Define what evidence will allow the route to return to ordinary practice.
Adaptation should not create permanent low expectations
Repeatedly reducing task complexity for the same learners can become a self-confirming pattern.
Use temporary supports with re-entry into the common goal wherever possible.
Adaptive decisions should be documented at the level needed for continuity
A useful note can be short:
Evidence → interpretation → adaptation → learner response → next check.
Example: “Correct method after model but not on unlabelled item → selection cue still carrying performance → mixed pair comparison → succeeds with one cue → next lesson fresh mixed set without cue.”
Do not average unlike evidence into one learner score
A learner can be strong in explanation and weak in retrieval, strong in routine execution and weak in transfer, or strong untimed and unstable under time.
Adapt to the component that matters rather than compressing all evidence into “ability.”
Common adaptive-teaching failure modes
- changing activities without changing the instructional problem;
- adapting from one answer without enough evidence;
- lowering the goal when a route change would suffice;
- using fixed learning-style labels;
- treating confidence as understanding;
- responding to all errors with more practice;
- responding to all errors with more explanation;
- keeping groups fixed after learner states change;
- letting the strongest learner define the pace for everyone;
- giving the entire class maximum support because a few learners need it;
- giving struggling learners easier work forever;
- extending strong learners only with more volume;
- removing accessibility supports in the name of independence;
- failing to test whether a temporary adaptation can be withdrawn;
- using AI recommendations as opaque diagnoses;
- recording adaptations but never checking their effect.
A practical adaptive-teaching sequence
- Define the learning goal and prerequisite assumptions.
- Plan evidence that can reveal meaningful states.
- Collect responses from enough learners to guide the decision.
- Preserve support conditions.
- Locate the first consequential breakdown.
- Distinguish knowledge, misconception, retrieval, execution, transfer and access barriers.
- Decide whether to pause and fix, adapt support, regroup, extend or gather more evidence.
- Keep the core goal intact where possible.
- Use the smallest useful route change.
- Let learners act under the changed condition.
- Collect fresh evidence.
- Fade temporary support when capability improves.
- Return learners to shared work where useful.
- Record only enough to support continuity.
- Recheck after delay or variation.
Verify that the adaptation changed learning, not only activity
A new worksheet, group or explanation is not evidence of success.
Use a fresh attempt under conditions that reveal whether the targeted barrier has changed.
If the learner succeeds only while the adaptation remains present, the next job may be fading rather than another adaptation.
Evidence, interpretation and limits
Current guidance supports checking understanding and using learner evidence to adjust teaching. EEF’s January and February 2026 Check. Adapt. materials explicitly focus on using responses to choose what to do next, while emphasising professional judgement rather than a formula. AERO’s Monitor Progress guide, last updated 14 May 2026, similarly frames checking for understanding as a basis for additional instruction, guidance or feedback.
Evidence for particular adaptive practices varies. Grouping, scaffolding, tutoring, feedback and formative assessment have distinct evidence bases and should not be bundled into one universal “adaptive teaching effect.” Adaptation quality depends on diagnosis, content knowledge, implementation and whether the changed route preserves the intended construct.
The worked cases and decision architecture in this article are editorial teaching designs. They have not been evaluated together as one intervention. No universal adaptation threshold, grouping rule or response percentage is claimed.
Sources for this edition were reviewed on 22 September 2026.
The adaptive-teaching standard: change the route because the evidence changed
Adaptation is not a performance of flexibility.
The teacher checks, interprets, chooses, acts and checks again.
Adaptive teaching succeeds when the next move becomes more appropriate to the learner’s actual state without shrinking worthwhile learning into a permanently easier destination.
Continue through How Teaching Works, revisit How Diagnostic Teaching Works, or use How Scaffolding and Fading Work in Teaching for the support architecture.
