How Teaching Assistant Interventions Work | Why Extra Adult Support Helps Only When Roles, Training and Handoffs Are Designed

A teaching assistant can make a classroom feel more supported and still leave the pupils who struggle most with less access to the teacher.

That sounds contradictory until you watch the route through a lesson. The teacher begins an explanation. A pupil hesitates. A well-meaning adult moves beside them, simplifies the task, supplies the first answer, keeps them moving and stays close for the rest of the period. The pupil completes more work. The room is calmer. Yet the learner has missed part of the teacher’s explanation, practised fewer independent decisions, and become increasingly dependent on the adult who was meant to make access easier.

This is why the phrase more adult support is too vague to describe an educational mechanism. Another adult in the room can increase feedback, provide structured practice, make an intervention possible, support communication, help a pupil re-enter a task or free the teacher to work differently. The same adult can also become an accidental barrier between a learner and high-quality teaching.

Current Education Endowment Foundation guidance makes that distinction unusually clear. Its Teaching Assistant Interventions evidence summary reports a positive average effect for targeted TA-delivered interventions, while the 2025 third edition of Deployment of Teaching Assistants warns schools not to use teaching assistants as substitutes for teachers for the pupils who need the strongest teaching. The updated guidance draws on a new review of more than 52 studies; the EEF Toolkit’s broader TA-intervention synthesis reports 65 studies and an average impact equivalent to about four additional months of progress, while rating the evidence moderate rather than definitive. EEF also stresses that structured, trained, targeted one-to-one or small-group work has a different evidence profile from general, loosely defined classroom deployment.

That difference is the centre of this article.

This page owns one narrow mechanism on eduKateSG: the design of teaching-assistant-delivered academic intervention and supplementary classroom support so that roles, training, scaffolding, communication, dosage and handoffs increase access to teaching rather than replace it. It does not own small-group tuition as a general teaching format, SEN plans, classroom accommodations, one-to-one tutoring, peer tutoring or ordinary teacher–student interaction. Those are neighbours.

The useful question is not, “Do we have an extra adult?” It is: What learning action becomes possible because this adult is here, and does that action leave the pupil more capable when the adult steps away?

The 50-second answer

Teaching assistant interventions work best when the extra adult has a defined instructional job rather than a general instruction to “help.”

A useful design sequence is:

1. Start with a diagnosed learning need. Know what the learner cannot yet do and what evidence would show improvement.
2. Protect access to the teacher. The pupil who struggles should not routinely receive less teacher explanation because the TA is nearby.
3. Use a structured intervention or clearly specified classroom role. Training, materials and routines should reduce improvisation.
4. Agree the scaffolding ladder. Prompt, clue, model, correct or temporarily support only as much as the pupil needs.
5. Make the pupil do the cognitive work. Support should preserve thinking rather than replace it.
6. Coordinate teacher and TA. Both adults need enough information about lesson goals, misconceptions and next steps.
7. Track attendance, dosage and progress. Offered support is not the same as received support.
8. Hand learning back to ordinary lessons. The classroom teacher should know what became secure and what remains fragile.
9. Fade support when independence grows. Permanent adult proximity is not the success condition.

The shortest principle is: extra adult time helps when it creates extra learning, not merely extra completion.

1. Begin with the learning problem, not the available adult

Schools can accidentally design support backwards. A teaching assistant has thirty minutes free, so a group is formed; only afterwards does anyone ask what the group should learn.

A stronger design starts with the pupil evidence. Perhaps six pupils can decode accurately but cannot read connected text fluently. Perhaps a learner understands ratio diagrams but cannot translate the relationship into symbolic form. Perhaps a group knows sentence punctuation but cannot expand short sentences with useful detail.

The learning problem determines the intervention, not the staffing timetable.

Concrete example: Instead of “TA maths group, Tuesday and Thursday,” the plan says, “Six-week ratio-representation intervention for five pupils who confuse part-to-part and part-to-whole relations; twelve 25-minute sessions; fresh transfer item each Friday.”

Now the adult, time and material belong to a mechanism.

Caveat: School staffing is never frictionless. Timetables constrain what is possible. The point is not to ignore staffing; it is to stop treating availability as a sufficient educational rationale.

2. Protect the learner’s relationship with the teacher

The EEF’s updated deployment guidance makes a particularly important equity point: pupils who struggle most should spend at least as much time with the teacher as their peers, not less.

That principle corrects a familiar pattern. The teacher explains to the class while the TA quietly re-explains to the same small group. Those pupils therefore hear neither explanation in full. Over months, the class learns from the teacher and the struggling pupils learn from a parallel adult system.

Concrete example: During whole-class modelling, the TA does not begin a side lesson. They help the pupil orient to the shared model, perhaps by pointing to the relevant line or quietly restating one instruction. The teacher remains the source of the core explanation.

Later, the TA may deliver a separate structured intervention at a time deliberately chosen not to remove the pupil from the most valuable whole-class input.

Caveat: Some specialist programmes legitimately require separate teaching. The rule is not “never leave the classroom.” It is “do not create a lower-expectation parallel curriculum by default.”

3. Distinguish general classroom support from structured intervention

The phrase teaching assistant intervention often hides two very different jobs.

One job is classroom deployment: supporting access, checking understanding, facilitating participation, helping routines function and scaffolding tasks.

The other is a defined intervention: a programme or protocol with a target group, sequence, dosage, materials and progress checks.

Evidence is more favourable when TAs are trained to deliver structured interventions than when they are simply placed beside lower-attaining pupils without a clear instructional design.

Concrete example: “Sit with Jayden in English” is a deployment instruction. “Deliver the 20-minute oral-language sequence three times weekly, using these prompts and recording these two measures” is an intervention design.

The difference is not bureaucracy. It changes what the adult knows to do when the pupil hesitates.

4. Train the instructional move, not only the resource

Handing a TA a worksheet is not training.

The adult needs to know what the task is intended to reveal, what a common wrong response means, which prompt should come first, when to model, when to wait, what not to tell the learner, and what counts as evidence that support can be reduced.

Concrete example: In a reading intervention, the TA learns that a particular hesitation should first trigger a decoding prompt rather than immediate word supply. They practise the prompt, see examples, rehearse common pupil responses and learn the stop rule for prolonged guessing.

This is different from “Here are the cards; work through them.”

Training should therefore include the mechanism behind the material.

Caveat: Training does not need to become a long credential for every small task. The amount of preparation should match the instructional complexity and risk.

5. Clarify who plans, who delivers and who owns the learning

Poor role definition creates hidden gaps.

If the teacher assumes the TA is planning the intervention and the TA assumes the teacher is, neither may have checked whether the work matches the curriculum. If the TA notices a misconception but has no route to report it, useful evidence disappears.

A simple role agreement can answer:

• Who selects pupils?
• Who identifies the target?
• Who prepares materials?
• Who delivers sessions?
• Who monitors attendance?
• Who reviews learning?
• Who communicates with families if needed?
• Who decides when support fades or changes?

The classroom teacher should retain educational ownership even when another adult delivers part of the support.

6. Make the lesson goal visible to both adults

A TA cannot scaffold toward a target they do not know.

“Help them with the worksheet” invites local completion. “Today they need to distinguish a cause from supporting detail” makes the cognitive goal visible.

Concrete example: Before class, the teacher gives the TA a thirty-second briefing: “The key thing is that pupils choose evidence that proves the claim. If they ask what to write, do not supply a sentence; ask which line in the source changes the answer.”

That tiny handoff changes the kind of help delivered.

A shared lesson goal also lets the TA notice whether a pupil’s difficulty is conceptual, linguistic, attentional or procedural.

7. Use a scaffolding ladder instead of one habitual kind of help

Adults develop favourite helping moves. Some ask questions. Some explain. Some point. Some give the first step. Some correct immediately.

A scaffolding ladder gives the adult several levels and a reason to choose among them.

A useful progression can move from lighter to heavier support:

1. pause and allow self-correction;
2. prompt the learner to recall the strategy;
3. point to a relevant cue or representation;
4. give a clue;
5. model one part;
6. provide direct correction where necessary;
7. return the task to the learner.

EEF’s 2025 TA guidance includes a scaffolding framework built around ideas such as correcting, modelling, clueing, prompting and self-scaffolding.

The goal is not always to start with the lightest possible hint. It is to use enough help while preserving a route back to independent action.

8. Avoid the answer-giving trap

The fastest way to increase task completion is often to tell the learner what to do.

That can make the session look successful while reducing learning.

Concrete example: A pupil cannot begin a word problem. The TA says, “It’s division. Write 48 ÷ 6.” The page moves. But the difficult decision—identifying the structure of the problem—has been performed by the adult.

A better scaffold might ask, “What are the quantities? Do we know the number of groups, the size of each group, or the total?” The pupil still has to select the relationship.

If the learner cannot, the TA may model one analogous example, then return to the original.

Completion is evidence only when the learner completed the important thinking.

9. Preserve productive struggle without manufacturing frustration

A TA can rescue too quickly because sitting beside a struggling child makes silence uncomfortable.

The adult needs permission to wait.

A five-second pause may allow retrieval. A prompt may allow the student to re-enter. A clue may be enough.

At the same time, “productive struggle” should not become a justification for making a pupil fail publicly for ten minutes.

Concrete example: The TA says, “Use the first step of the routine,” then waits. The pupil begins. No further help is needed.

Caveat: Learners differ. Processing, language, anxiety, disability and prior knowledge change how much time and support are reasonable. The goal is not maximal struggle. It is maximal useful learner action.

10. Ask questions that preserve the problem

A question can scaffold or it can secretly contain the answer.

“Is the answer 12?” is an answer with a question mark.

“What relationship do these two quantities have?” keeps the problem alive.

“What did you do last time?” may support strategy retrieval.

“Where does your working stop making sense?” invites diagnosis.

Good TA questioning should make the learner’s reasoning more visible without turning the interaction into a rapid-fire interrogation.

A useful test is: after the question, who still has to make the important decision?

11. Build subject knowledge into support

General warmth and patience matter. They are not enough for every academic intervention.

A TA supporting fractions needs enough mathematical content knowledge to avoid reinforcing misconceptions. A TA supporting phonics needs accurate knowledge of the code and the programme’s instructional sequence. A TA helping analytical writing needs to distinguish evidence from explanation.

This does not mean every TA must be a subject specialist. It means leaders should match instructional complexity with training and supervision.

Caveat: When a question exceeds the adult’s confidence, “Let’s flag this for your teacher” is often better than improvised teaching.

12. Use structured programmes without turning adults into scripts

A structured intervention reduces harmful variation. It gives the TA a sequence, examples, prompts, dosage and decision rules.

But fidelity is not identical to robotic delivery.

A learner may need a repeated example, a language clarification or a slower pace. The adult still needs judgment.

Concrete example: The programme says pupils should answer four items independently after modelling. One pupil’s first two errors show the prerequisite concept is missing. The TA uses the programme’s repair route rather than mechanically completing all four.

The active ingredients should be protected while surface delivery remains humane.

13. Protect dosage from timetable erosion

An intervention scheduled three times a week may quietly become once a week because assemblies, trips, cover, transport and competing duties consume the slot.

This is an implementation failure, not evidence that the intervention “did not work.”

Track received dosage:

• sessions offered;
• sessions attended;
• minutes actually delivered;
• major interruptions;
• staff changes.

Concrete example: A six-week programme was designed for eighteen sessions. One pupil received eleven. Their result should not be interpreted as a full-dose test of the programme.

Caveat: More minutes are not automatically better. The target is sufficient, high-quality exposure.

14. Keep groups instructionally coherent

Small groups are efficient when pupils share a teachable need.

A group assembled because all six pupils are “below expected standard” may contain six different bottlenecks.

Concrete example: Two pupils need decoding repair, two need fluency, one needs vocabulary and one understands text but struggles to express answers. A single generic reading group will dilute the instructional target.

Group from diagnostic evidence, not only attainment band.

This principle links naturally to flexible grouping: when the data change, the group should change.

15. Do not let TA support become permanent ability grouping

Once a pupil enters “the TA group,” institutional gravity can keep them there.

The adult relationship becomes stable; expectations become lower; the curriculum narrows; peers and pupil begin to treat the placement as identity.

Every targeted intervention needs an exit condition.

Concrete example: “Exit after two consecutive fresh probes show accurate use of the taught inference routine in ordinary class reading,” not “continue until the end of term because that is the group.”

Support should be revisable evidence, not a label.

16. Plan the handoff back into ordinary teaching

An intervention can succeed in a quiet room and vanish in the classroom.

Transfer needs design.

The TA and teacher should know how the intervention language maps onto ordinary lessons. The learner should meet the same cue, representation or strategy again when appropriate.

Concrete example: A TA teaches a “say the gist in one sentence” routine. The English teacher uses the same language during whole-class reading later that week. The pupil is prompted once, then expected to use it independently.

The intervention has entered the mainstream learning environment.

17. Communicate progress in instructional language

“Had a good session” is kind but not useful enough.

A handoff should say what changed.

Useful notes can be short:

• target;
• evidence secure;
• error still present;
• prompt that worked;
• next transfer task.

Example: “Can now compare fractions with common numerator using magnitude reasoning without the strip. Still reverses comparison when denominators differ and numerator is close. Number-line prompt helped; try one fresh item in class tomorrow.”

This is evidence the teacher can act on.

18. Prevent the TA from becoming the pupil’s executive-function system

A nearby adult can remember the equipment, interpret every instruction, initiate every task, monitor every deadline and notice every mistake.

The pupil appears organised because the adult is doing the organising.

Scaffolding should gradually move these functions back to the learner.

Concrete example: Instead of saying, “Get your book out; write the date; start number one,” the TA points once to the student’s checklist and waits.

Later the checklist remains while the adult moves away.

Independence is not abandonment. It is a planned redistribution of control.

19. Use proximity strategically

Constant proximity communicates support but can also reduce autonomy, peer interaction and teacher access.

A TA can position themselves so the learner can seek help without being visually marked as permanently assisted.

Concrete example: During independent work, the TA checks a cluster of pupils, including the target learner, then moves elsewhere. They return only if the agreed signal appears.

This makes support available rather than compulsory.

Caveat: Some safety, medical or access needs genuinely require close adult presence. The educational aim should still be the greatest appropriate independence.

20. Keep peer relationships open

One-to-one adult support can unintentionally isolate a pupil from classmates.

If the TA answers every question, joins every group and interprets every social interaction, peers stop learning how to work with the learner directly.

Concrete example: During a collaborative task, the TA prepares the pupil to contribute one idea, then steps back so peers respond to the pupil rather than to the adult.

The adult supports participation without becoming the social gateway.

21. Avoid translating every teacher instruction into simpler language

Restating can help when language genuinely blocks access. Doing it automatically can teach the learner not to listen to the teacher because a private version will follow.

A better approach is selective.

First check whether the pupil understood. If not, clarify the smallest blocking element.

Concrete example: Instead of repeating a six-step instruction, the TA asks, “What is the first thing you need to do?” The pupil answers correctly and starts.

No translation was needed.

This is another reason the TA must know the lesson goal rather than simply shadow the pupil.

22. Support metacognition without replacing it

A TA can make learning decisions visible:

• What is the task asking?
• What strategy fits?
• What evidence do you have?
• Where did the error begin?
• What will you check before asking for help?

These prompts help the pupil internalise a control routine.

But a student who is asked the same metacognitive questions every lesson may become dependent on the adult cue.

Fade the prompts and see whether the learner asks them internally.

23. Use correction that returns quickly to successful practice

When a pupil makes an error, the TA’s job is not to deliver a miniature lecture unless the misconception demands one.

Efficient correction identifies the error, supplies or elicits the needed information, and returns the learner to a successful attempt.

Concrete example: In a spelling intervention, the pupil omits a phoneme. The TA segments the word, helps map the missing sound, then asks the pupil to spell the word again and use it in a short sentence.

The correction ends with learner production.

24. Separate emotional support from avoidance of challenge

A trusted TA can become the person who helps a pupil escape difficult work.

Warmth and high expectations have to coexist.

Concrete example: The learner says, “I can’t do this.” The TA acknowledges the frustration, reduces the task to the first decision, and stays long enough for the pupil to begin. They do not substitute an easier worksheet merely to restore calm.

Caveat: Sometimes reducing demand is appropriate because the task is inaccessible, the learner is dysregulated or the target has been misjudged. The distinction requires professional knowledge, not toughness.

25. Make accommodations visible to the team

A TA may know the practical details of what helps a pupil—where to sit, how to chunk instructions, which visual cue works—while the teacher sees only the final performance.

That knowledge should not remain private.

A simple team record can distinguish:

• required accommodation;
• temporary scaffold;
• successful prompt;
• independent capability.

This prevents one adult from becoming the only person who knows how the learner can succeed.

26. Use family communication carefully

Parents may assume that more TA time means more learning. Schools should explain the actual role.

If a structured intervention is occurring, families can be told the target, schedule and review point. If the TA supports classroom access, explain how the support protects teacher contact and independence.

Avoid reporting only that the child “has support.”

The meaningful information is what the support is designed to change.

Caveat: Communication should respect privacy, school policy and the roles of teacher, SENCO and other professionals.

27. Watch for hidden curriculum narrowing

Students supported by TAs sometimes complete less cognitively demanding versions of classwork.

The intention is access; the result can be lower opportunity to learn.

Concrete example: The class analyses a source while the supported pupil matches vocabulary definitions. The vocabulary task may be useful, but if it permanently replaces the central disciplinary thinking, the pupil is being protected from the curriculum.

A better support might pre-teach two blocking terms and then help the learner join the same source-analysis question with a scaffold.

Support should preserve the target whenever possible.

28. Let the teacher learn from the TA’s observations

TAs often see micro-patterns the teacher cannot see while managing the whole room.

A pupil may always stall at the second instruction, misread a particular symbol, copy a peer before asking for help or succeed when the diagram is visible.

Those observations are valuable when translated into testable instructional information.

Concrete example: “She can solve the equation once it is written symbolically; the breakdown is translating the sentence into the equation.”

That observation may change tomorrow’s whole-class teaching.

The TA should not become merely an extra pair of hands. They can be an extra source of instructional evidence.

29. Give TAs a route to say “I don’t know”

Poor systems create pressure for adults to answer every pupil question.

A healthy system permits uncertainty.

“I’m not certain this is the method your teacher wants. Let’s mark the question and check.”

That response protects accuracy and models epistemic responsibility.

The teacher can later clarify the method for both adult and pupil.

Caveat: Frequent uncertainty on core intervention content signals a training problem that leaders should fix.

30. Review whether the intervention is worth its opportunity cost

TA time is finite.

A 25-minute intervention costs not only salary but also the other learning, supervision or collaboration the adult could have supported.

Review:

• learning gained;
• attendance;
• staff preparation time;
• pupil experience;
• teacher access preserved or lost;
• transfer;
• independence.

An intervention with small gains may still be worthwhile if the alternative is weaker. A fashionable programme with weak fit may not be.

Cost-effectiveness is a local decision informed by evidence, not a universal league table.

31. Fade the adult before you declare independence

The pupil may succeed because the same adult, seat, prompt and timing are still present.

Remove support in stages.

Concrete example: Week one: TA checks every task start. Week two: checks only the first task. Week three: pupil uses a self-check card and the TA observes from a distance. Week four: no planned check unless the learner uses the help signal.

If performance collapses, the fade was too fast or the capability was not secure.

The test of support is what survives when support becomes lighter.

32. Escalate when low-intensity support is not enough

Teaching assistants are not a substitute for specialist assessment, qualified teaching, safeguarding processes, speech and language therapy, educational psychology or other professional provision.

Repeated failure despite good implementation is information.

If a pupil’s need remains persistent, severe or puzzling, the school should widen the response rather than endlessly extend the same TA support.

The strongest systems know what the intervention can do and where its competence ends.

Worked case 1 — The pupil who received less teacher because she had more support

A Year 7 pupil with weak reading comprehension sat beside a TA in every English lesson. The adult whispered explanations, rephrased questions and helped compose answers. The pupil completed most tasks but rarely answered the teacher directly. Video observation showed that she spent almost a quarter of whole-class explanation time listening to the TA instead of the teacher.

The team changed the design. Whole-class explanation became protected teacher time. The TA used a visual cue rather than parallel speech and noted points of confusion. Twice weekly, the pupil attended a separate structured comprehension intervention. During class practice, the TA visited briefly and then moved away.

The quantity of adult attention did not increase. Its architecture changed. Within several weeks, the pupil was answering more teacher questions and needed fewer private re-explanations.

Worked case 2 — The intervention with excellent materials and poor dosage

A school purchased a literacy programme with strong training materials. The nominated TA was capable and enthusiastic. Yet assemblies, staff cover and school events cancelled nine of eighteen planned sessions.

At review, progress was modest. The first interpretation was that the programme had failed.

Attendance logs changed the diagnosis. Pupils had received roughly half the planned exposure. The next cycle protected the intervention slot, assigned cover for the TA’s other duty and created a simple missed-session recovery rule.

The second group received the intended dose and showed clearer gains.

The lesson is not that dosage guarantees impact. It is that implementation evidence must accompany outcome evidence.

Worked case 3 — The helper who made the learner helpless

A Primary pupil had difficulty beginning independent writing. The TA, wanting to prevent distress, suggested the opening sentence each day. The pupil became quicker to start when the TA was present and slower when the adult was absent.

The team identified the hidden dependency. They replaced sentence supply with a three-step start card: choose the subject, say the first idea aloud, write a plain first sentence. The TA initially prompted each step, then pointed only to the card, then stopped prompting.

The pupil’s early sentences were less polished than the adult-generated versions. They were also the pupil’s.

Independence improved because the intervention stopped confusing product quality with learner control.

Worked case 4 — The small group that contained four different problems

Six pupils were placed in a “reading support” group because their comprehension scores were low. Sessions were generic.

Closer diagnostic work showed different bottlenecks: two pupils had decoding weaknesses, one had slow fluency, two lacked vocabulary and one could understand orally but struggled to write answers.

The school dissolved the group. The first two entered a decoding intervention, one received fluency practice, two joined vocabulary-rich guided reading, and the final pupil received written-expression support while staying in grade-level reading.

The staffing became slightly more complex. The teaching became much more coherent.

Practical route for teachers

Before the lesson, decide what the TA should help pupils do, not merely whom the TA should sit beside. Share the learning goal, likely misconception and the lightest useful prompt. Protect whole-class teacher explanation. During practice, let the TA gather evidence rather than complete tasks for pupils. After class, use a thirty-second handoff: what became secure, where did reasoning break, and what should tomorrow’s teacher move be?

For structured interventions, agree the target, group, schedule, training, progress measure and exit condition before the first session. Check received dosage. Observe at least one session. Make sure the language and representations connect back to ordinary lessons.

Most importantly, audit dependency. If the learner succeeds only beside the adult, the support has not yet completed its job.

Practical route for learners

Know what the adult is helping you learn. Ask, “What is the strategy I am supposed to be able to use by myself?” When you are stuck, try the agreed first step before asking for an answer. If a prompt helps, notice what kind of prompt it was so you can reproduce it later.

When the TA steps away, treat that as part of the intervention rather than withdrawal of care. Independence is evidence that the support is working.

If the work becomes impossible when support fades, say exactly where it breaks. That information helps the teacher distinguish a missing skill from a missing reminder.

Practical route for parents and families

Ask the school three concrete questions: What is the support targeting? How will you know it helped? What is the route toward greater independence?

More adult time is not automatically better. A good plan protects your child’s access to qualified teaching, uses the TA for a defined job, and reviews whether the child can perform more of the task without constant adult presence.

At home, avoid turning the TA into the child’s only trusted route into schoolwork. Encourage the child to explain the strategy they are learning and to bring unresolved questions back to the teacher.

Common failure modes

• Assigning a TA to “help” without defining the learning target.
• Allowing the pupil with the greatest need to receive the least teacher contact.
• Running side explanations during whole-class instruction.
• Giving worksheets instead of training.
• Using one favourite prompt for every difficulty.
• Supplying answers because completion looks like success.
• Creating groups from attainment bands rather than shared bottlenecks.
• Cancelling intervention sessions repeatedly and still judging the programme at full strength.
• Keeping pupils in support groups after the target is secure.
• Failing to hand intervention learning back to the classroom teacher.
• Letting the TA become the learner’s memory, planner and task initiator.
• Treating adult proximity as inherently supportive.
• Narrowing the curriculum for supported pupils.
• Asking families to interpret “has TA support” as a description of the educational mechanism.
• Keeping useful TA observations informal and invisible to the teacher.
• Improvising subject explanations outside the adult’s training.
• Treating a structured programme as a rigid script.
• Fading support so quickly that success collapses.
• Never fading support at all.
• Using low-intensity TA support when a specialist response is required.

Frequently asked questions

Do teaching assistants improve attainment?

Targeted, structured TA-delivered interventions have a positive average effect in the EEF evidence base, but general deployment is much more variable. The adult’s role, training, programme structure, pupil attendance and connection to teacher-led instruction matter.

What does the EEF currently report?

The EEF Teaching Assistant Interventions Toolkit reports an average impact of about four additional months of progress, moderate cost and moderate evidence, based on 65 studies. Its 2025 deployment guidance uses a newer review of more than 52 studies to make system-level recommendations about effective TA use.

Should the lowest-attaining pupils spend most of the lesson with the TA?

Not by default. The 2025 EEF guidance explicitly argues that pupils who struggle most should have at least as much access to the teacher as others, rather than being routinely separated from the strongest classroom teaching.

Are TAs effective in one-to-one work?

They can be, especially when trained and supported to deliver a structured intervention. One-to-one format alone does not guarantee quality.

Is a TA the same as a tutor?

No. A tutor is an instructional role defined by teaching a learner or group. A TA may deliver tutoring-like intervention, but also supports classroom access, routines, scaffolding and teacher workload. The job title does not describe the mechanism.

How large should a TA intervention group be?

There is no universal optimum. The group should be small enough for the programme to deliver its intended feedback and practice, and coherent enough that pupils share the target need.

Should the TA always simplify work?

No. Simplification can sometimes create access, but unnecessary simplification can lower the learning target. Preserve the same disciplinary thinking where possible and change the scaffold, representation, pacing or support.

How much should the teacher and TA communicate?

Enough to align goals, methods, evidence and next steps. This can be brief when systems are well designed. A thirty-second pre-lesson briefing and concise post-session note can be more valuable than long but irregular meetings.

When should support be faded?

When the learner can perform the target with reduced prompting across relevant contexts. Fade gradually and watch whether performance survives.

What if progress remains weak?

Check diagnosis, attendance, dosage, instructional quality, group fit, access to teacher teaching and the suitability of the intervention. If need remains persistent or complex, escalate rather than repeating the same support indefinitely.

Sources and further reading

• Education Endowment Foundation, Teaching Assistant Interventions: https://educationendowmentfoundation.org.uk/education-evidence/teaching-learning-toolkit/teaching-assistant-interventions
• Education Endowment Foundation, Deployment of Teaching Assistants, third edition, 26 March 2025: https://educationendowmentfoundation.org.uk/education-evidence/guidance-reports/teaching-assistants
• Education Endowment Foundation, What goes up, must come down: Promoting pupil independence through scaffolding, 22 April 2025: https://educationendowmentfoundation.org.uk/news/what-goes-up-must-come-down-promoting-pupil-independence-through-scaffolding
• Education Endowment Foundation, Selecting interventions for pupils with SEND: https://educationendowmentfoundation.org.uk/education-evidence/guidance-reports/selecting-interventions

Continue exploring on eduKateSG

• How X Works Hub
• How Small-Group Tuition Works for Performance
• How Explicit Instruction Works
• How Guided Practice Works
• How Classroom Accommodations Work

The final idea

A teaching assistant is not an intervention simply by being present.

The educational value comes from architecture: the learner still receives strong teacher teaching; the additional adult has a defined job; the adult knows what success and error look like; support preserves thinking; the intervention receives enough dosage; useful evidence returns to the teacher; and the scaffolding gets lighter as the pupil becomes more capable.

That standard is demanding because it removes a comforting shortcut. Schools cannot measure support by adult proximity alone.

The right question is harder and more useful: What can the learner now do because this support existed—and can they still do it when the support steps back?