How Questioning Works in Teaching | Designing Questions That Reveal Thinking and Guide the Next Move

Three students give the correct answer. The teacher is about to move on.

In this imagined classroom, the question is whether three-quarters is greater than three-eighths. One student compares three pieces of two different sizes. Another remembers a rule about the smaller denominator, but has forgotten when that rule applies. A third changes an answer after noticing a neighbour’s board.

The visible result is agreement. The learning underneath it is not the same.

Now the teacher asks a second question: “Would one-quarter also be greater than three-eighths? Show how you decided.” The first student can reconstruct the comparison. The second encounters the boundary of the remembered rule. The third must produce an answer without borrowing the first response. One additional question has made the next teaching decision more informed.

Questioning works in teaching when a question creates useful thinking, makes some of that thinking observable, and helps the teacher choose an appropriate next move. The quality of the exchange depends on the question, the conditions for answering, the interpretation of the response and what happens afterwards. Asking more questions is not the same as knowing more about the learner.

This guide continues the How Teaching Works series after How Worked Examples Work in Teaching. Its focus is the teacher’s design and decision-making work. For students learning to create their own study questions, the companion is How Question Generation Works in Learning.

The classroom scenes, passages, numbers and response sequences below are original illustrative examples, not records of actual students, experimental findings or official examination questions. They show ways to put the principles into practice; they are not a validated diagnostic instrument or a promise of particular results.

A route through this guide

Begin with the relationship between a question and a decision, then examine question design, response conditions and interpretation. The worked teaching cases cover Mathematics, English and Science data. The later sections address discussion, lesson planning, a reusable question bank and the evidence and its limits.

A question is only as useful as the decision it can improve

Before writing a question, name the decision it is supposed to inform. Should the lesson move on? Does a prerequisite need attention? Should the teacher demonstrate another example or withdraw a prompt? Is the learner choosing the wrong method, applying the right method inaccurately, or misunderstanding the demand?

These are different decisions. A question that answers one may tell the teacher little about another. Asking for the definition of a term can check access to that definition, but it does not by itself establish whether the learner can use the term to interpret unfamiliar evidence. Asking for an explanation can expose reasoning, but an elaborate explanation may still contain a factual error. The response must match the decision being made.

This is consistent with the Australian Education Research Organisation’s guidance on monitoring progress: checking understanding should help teachers identify what students can do and respond with additional instruction, guidance or feedback where needed. The useful connection is between evidence and instructional action, not between the number of questions asked and an assumed level of teaching quality.

A practical planning sentence is: “I am asking this because the answer will help me decide whether to ______.” Complete that sentence before choosing the wording. If every possible response leads to exactly the same lesson, the question may still have a useful practice or discussion purpose, but it is not doing much diagnostic work.

Questioning also has a cost. Each exchange uses time, attention and classroom space. There is no need to interrupt every sentence of an explanation. Choose points where misunderstanding would matter: a new distinction, the first use of a method, a change in representation, or the moment before learners begin independent work. A well-placed question can protect the next ten minutes from being spent on the wrong task.

Different questions do different intellectual jobs

It is tempting to divide questions into simple and difficult, closed and open, or lower-order and higher-order. Those descriptions can help organise a lesson, but none tells the teacher whether a particular question is well chosen. A short question can expose a deep misconception. An open question can invite a long answer that avoids the central issue.

A retrieval question asks learners to bring something to mind: a term, relationship, event, procedure or principle. It is useful when the next task depends on that knowledge being accessible. “What does the denominator tell us?” is not a lesser question when denominator meaning is the missing prerequisite.

An explanatory question asks for a connection. “Why does applying the same operation to both sides preserve the equality?” requires more than naming an operation. “How does this detail support your interpretation?” requires a link between evidence and claim. The IES instruction and study practice guide includes explanatory questioning among its recommendations. That supports deliberately asking learners to explain relationships; it does not make every question beginning with “why” equally effective.

A discrimination question asks learners to separate neighbouring possibilities. “Why does this problem need this method rather than that one?” checks selection. “Which of these statements is an observation, and which is an inference?” checks a boundary. Such questions are useful when learners know individual ideas but confuse their conditions of use.

A prediction question asks what should follow from a model. A prediction becomes more informative when the learner also gives a reason. A correct guess is not equivalent to an explanation that would generate further correct predictions. The teacher can compare the predicted result with what the model or evidence actually permits.

A transfer question changes something about the original task and asks whether the relevant learning can still be used. A reflective question asks learners to examine their decisions: “Which step did you check?” or “What evidence changed your mind?” Reflection should concern something inspectable, rather than invite a vague account of having worked hard.

A lesson may need all of these jobs, but not at every moment. The sequence should follow the learning problem. First establish the knowledge required to participate, then ask learners to explain, distinguish, apply and review it. When a response exposes a gap, return to the appropriate earlier step rather than increasing the sophistication of the question while the foundation remains absent.

Design the evidence before polishing the wording

Consider the question, “Do you understand equivalent fractions?” It does not specify what the learner should produce. A yes might mean confidence, recognition, politeness or genuine understanding. A no might mean one particular step is unclear. Neither answer localises the problem well.

Now consider: “Write a fraction equal to three-quarters with a denominator of eight, and explain what changed.” The learner must produce an equivalent fraction and account for the transformation. The teacher can inspect whether the numerator and denominator were both adjusted, whether equality was preserved, and whether the explanation concerns value rather than merely a memorised operation.

The improvement is not that the second question sounds more academic. It produces evidence that is better aligned with the intended understanding. The teacher knows what to look for and can decide what to ask next. This is the question-design application of the broader approach in How Diagnostic Teaching Works.

Before using an important question, write a strong response and two plausible incomplete responses. Then ask what you would do with each. If you cannot distinguish the responses without inventing a hidden motive, revise the question or plan a follow-up. If several interpretations of your wording would be reasonable, clarify the wording before treating one interpretation as a student error.

For numerical questions, solve the task yourself and check the alternatives. For a text-based question, check that the requested inference is supported by the supplied text. For an explanation, identify the indispensable relationship rather than deciding that a response is good merely because it contains familiar vocabulary. A question bank needs answer analysis, not just an answer key.

Keep the demand manageable. “Describe the pattern, explain the cause, evaluate the experiment and suggest an improvement” contains several tasks. That may be appropriate for a developed written response, but it is a poor first diagnostic prompt when the teacher needs to know whether the learner can read the pattern at all. Separate the demands, then reconnect them when the components are secure enough.

Closed questions can expose reasoning; open questions can conceal it

A multiple-choice question is not inherently shallow. A carefully selected set of alternatives can ask learners to discriminate between similar claims, identify a valid transformation, or recognise which conclusion the evidence supports. Its advantage is that the teacher can compare responses quickly. Its limitation is that choosing an option does not reveal the whole route used to choose it.

A useful two-part design asks for both a choice and a reason. The reason can be a short sentence, an annotation, a calculation, a selected justification or a diagram. The format should expose the relevant thinking without demanding an unnecessary essay. A learner who chooses correctly for an invalid reason gives the teacher different evidence from a learner who can justify the choice.

Incorrect alternatives should be plausible for a reason, not simply absurd. However, do not assume that each option identifies one unique misconception. The same option can arise from misreading, guessing, a calculation error or a conceptual misunderstanding. An alternative is a prompt for investigation, not a psychological label attached to everyone who selects it.

Open questions require the same care. “Tell me everything you know about energy” may produce a long inventory without showing whether a learner can explain the situation at hand. “What changes in this situation, and what evidence tells you?” is narrower but can require more relevant reasoning. Breadth is useful when exploring possibilities; precision is useful when a particular decision must be made.

The most productive choice is often a sequence rather than a favourite format: a short independent answer, a request for reasoning, a comparison with an alternative, and a fresh application. Each step adds a different kind of information. The teacher can stop when the available evidence is sufficient for the next reasonable action.

Place a question where the lesson genuinely can change direction

A hinge question sits at a decision point. The teacher is about to introduce more complexity, reduce support or begin independent practice, and wants evidence about readiness. Its value lies in the planned response to the answers. A beautifully written question is not functioning as a hinge if the teacher is committed to moving on whatever happens.

For example, before moving from expanding expressions to solving equations that contain brackets, ask learners to expand a fresh expression and justify the multiplication. If a repeated error shows that multiplication has been applied to only one term, that prerequisite needs attention. If the expansion is secure but the arithmetic is unstable, the next task should address a different issue.

AERO’s formative assessment guidance recommends identifying important points in learning progressions, using low-key checks and asking students to articulate their reasoning. In practical planning, this means having a response ready: proceed, provide a brief clarification, reteach a prerequisite, or give a targeted task to learners who need it.

A percentage alone is not a universal decision rule. The consequences of moving on matter. One learner’s unresolved difficulty may need an individual follow-up; a widespread foundational error may make whole-class reteaching necessary. A correct class total can also hide the fact that the same small group repeatedly lacks access. Look at who answered, how they answered and which understanding the next task requires.

Plan a re-entry route for learners who need extra teaching. The lesson should not leave them permanently outside the main work. A targeted explanation followed by a fresh check can provide a practical route back. Equally, learners who are ready need a worthwhile extension, not simply more waiting while the teacher repeats material they can already use.

The conditions for answering are part of the question

The written wording does not fully define a classroom question. Was the model still on the board? Did a peer answer first? Was a formula provided? Were students asked to write independently or reach a group agreement? Did the teacher narrow the choice through tone, pointing or repeated hints?

These conditions change what the answer can tell us. Completing a response after a useful hint can be genuine progress. It is simply not the same observation as producing the response before help. The distinction is developed in A Supported Answer Is Not the Same Measurement as an Independent Answer.

When the goal is to see each learner’s starting point, collect a first response before discussion. A short note, a diagram, an answer card or a private digital response can work. Then discussion can begin. Preserve the difference between the first attempt and the revised answer so learning through discussion remains visible rather than being confused with prior knowledge.

AERO’s classroom demonstrations of monitoring progress illustrate varied participation routines, answer boards, constructive responses and follow-up teaching. These are examples of practice, not proof that a particular device or routine is always best. The practical aim is to hear from beyond the quickest volunteers and to give responses an instructional consequence.

Whole-class responses need to be readable. Asking everyone to hold up a tiny sentence that the teacher cannot inspect produces the appearance of participation without much information. Use a response length and format that can actually be examined. In a small group, a short explanation from each learner may be feasible. In a larger class, a brief common response followed by selected follow-ups may be more workable.

Do not turn every exchange into surveillance. Some questions exist to explore, rehearse or enjoy an idea together. The teacher should know when the purpose is a clean individual check and when it is collaborative thinking. Naming that distinction helps students understand why they sometimes work alone first and sometimes think together from the beginning.

Give learners time to form an answer, not just permission to speak

A question followed immediately by the teacher’s own answer is mainly an explanation in question form. That can be useful when intentionally modelling reasoning, but it provides little evidence about what learners could have produced. Decide whether you are demonstrating or eliciting; the classroom needs both, and confusing them weakens interpretation.

Mary Budd Rowe’s early wait-time research report examined pauses both after teacher questions and after student responses. Its reported findings linked longer pauses in the studied elementary science settings with changes in the amount and character of student contributions. This is a reason to examine conversational pacing, not a universal law that every learner needs exactly the same number of seconds.

Match thinking time to the task. Retrieving a familiar term differs from inspecting a graph, generating a counterexample or planning an interpretation. Reading demands, unfamiliar vocabulary and the response format also matter. “Take a moment to write your first idea” gives a clearer working invitation than silence whose purpose students do not understand.

The pause after an answer can matter too. A learner may have another sentence to add, or another student may be able to build on the reasoning. Immediately saying “correct” and moving on can close that opportunity. Sometimes a neutral follow-up is more useful: “What supports that?” or “Can anyone connect that step to the previous one?”

Waiting should not become abandonment. If a student has no viable entry point, prolonged public silence may reveal little and feel punitive. Offer a smaller demand, clarify a term, allow a written response, or teach the missing prerequisite. The aim is productive time for thinking, not a contest over how long the teacher can withhold help.

Make the reasoning demanding and the route into the question accessible

Difficulty with a response format is not automatically difficulty with the concept. A learner may be able to show a relationship with a diagram while struggling to formulate a public spoken answer. Another may understand an explanation but need the written question visible because the spoken wording was difficult to retain. These possibilities call for investigation, not assumptions about ability.

CAST’s guidance on multiple media for communication encourages varied ways of expressing knowledge when a particular medium is not itself essential to the learning goal. Applied to questioning, the important distinction is between the knowledge being examined and an avoidable barrier in how the learner is required to display it.

If the goal is mathematical comparison, a labelled drawing may offer useful evidence alongside words. If the goal is oral presentation, speaking is part of the intended performance and should be supported and assessed accordingly. If the goal is written argument, an oral explanation can help diagnose understanding, but the teacher still needs a later written attempt to evaluate writing.

Independence does not mean removing necessary accessibility supports. A student can reason independently while using an appropriate access tool or an agreed response arrangement. Record which instructional hints supplied parts of the answer; do not treat all assistance with access as equivalent to someone else doing the reasoning.

Questioning should also preserve dignity. Explain participation routines in advance, allow preparation, respond respectfully to uncertainty and avoid using an unexpected public question to expose or embarrass a student. Teachers can maintain high expectations while offering ways to enter the exchange. Accountability is compatible with care; humiliation is not a teaching objective.

Treat an answer as evidence, not direct access to a mind

The teacher observes words, marks, gestures and actions. Understanding is inferred from those observations. A single answer rarely establishes the entire cause of a difficulty. “This response suggests a denominator misconception” is more defensible than “This answer proves the learner does not understand fractions.”

Keep three things separate: what happened, what might explain it, and what should be checked next. For example: the learner changed only one term when expanding an expression; a distributive-property difficulty is one possible explanation; a fresh expression with simpler numbers can help test that possibility. This language is precise without becoming a permanent label.

Correct answers also need interpretation. A correct result with an unsupported reason should not be recorded as the same evidence as a correct result with a coherent justification. A correct explanation that depends on a supplied model can show progress without yet establishing independent use. A fluent spoken response can contain a hidden error. Confidence and speed are observations, not substitutes for checking the substance.

A useful follow-up changes one feature at a time when practical. Change the numbers but preserve the structure. Keep the concept but simplify the language. Keep the wording but remove a method cue. These variations do not create a perfectly controlled experiment, but they can reduce some of the ambiguity in the original response.

Interpret patterns across attempts rather than elevating every slip into a diagnosis. When the same error survives clarification and appears in several related tasks, the case for targeted teaching becomes stronger. When it disappears immediately after the learner re-reads a condition, the original response may not justify reteaching the whole concept. The decision should remain proportionate to the evidence.

Follow up without quietly answering your own question

There is a difference between a probing question and a leading hint. “How did you decide?” invites the learner’s route. “Shouldn’t you multiply the second term as well?” supplies a substantial part of the correction. Both may be appropriate at different moments, but they produce different kinds of evidence.

Start with a neutral invitation when a learner has something to explain. Ask which information was used, what a symbol means, or where a particular claim came from. If the response remains unclear, narrow the question. If the learner lacks the knowledge required to proceed, explain or model it. Do not prolong a guessing game in the hope that enough small hints will eventually resemble understanding.

Repeatedly asking “Are you sure?” is particularly ambiguous. It can communicate that an answer is wrong without revealing why. Learners may learn to reverse their answer whenever the teacher asks it. A more informative alternative names the checking job: “Test your value in the original equation” or “Which sentence supports that claim?” The learner is then asked to perform a check rather than read the teacher’s expression.

Once a hint has helped, acknowledge the progress and arrange a fresh attempt. The earlier supported response remains part of learning. The fresh task answers a new question: can the learner now make the decision without that particular hint? This is how questioning connects with the selective support and withdrawal discussed in How Explanation Works in Teaching.

Worked teaching case: the correct fraction answer with three different reasons

The learning goal in this original example is not simply to select the larger fraction. It is to compare fractions using their values and to recognise the conditions under which a shortcut is valid. The teacher begins with a manageable case, then changes a feature that makes an overgeneralised rule fail.

First question: collect an answer and its basis

Ask: “Which is greater, three-quarters or three-eighths? Explain with a drawing, equivalent fractions or a sentence about the sizes of the parts.” When using drawings of quantities, make clear that the wholes being compared are equal in size. Give each learner space to respond before showing another student’s answer.

A strong explanation might say that three-quarters equals six-eighths, which is greater than three-eighths. Another valid explanation compares three quarter-sized parts with three eighth-sized parts of an equal whole. The teacher should accept more than one sound route rather than require a particular sentence to be repeated.

Suppose a learner writes, “Three-quarters, because the smaller denominator always makes the fraction bigger.” The result is correct, but the word “always” creates a question about the rule’s boundary. It would be premature either to award complete conceptual confidence or to declare the learner unable to compare fractions. A second question can make the issue inspectable.

Second question: change the feature that tests the rule

Ask: “Now compare one-quarter and three-eighths. Does the same explanation still work?” Here one-quarter equals two-eighths, so three-eighths is larger. A rule based only on selecting the smaller denominator gives the wrong result. The example separates knowing a useful same-numerator comparison from applying it indiscriminately.

If the learner can produce two-eighths and three-eighths but struggles to phrase the conclusion, help with expression while preserving the mathematical evidence. If the learner changes the denominator without changing the numerator, equivalent-fraction knowledge needs closer inspection. If the learner compares correctly and explains the restriction, move towards a new application rather than repeating the original explanation.

Notice that none of these branches is determined by the final answer alone. The explanation and the intermediate representation help the teacher choose a response. Even then, the working interpretation remains provisional. One pair of questions is useful classroom evidence, not a complete assessment of the learner’s fraction understanding.

Teach the missing relationship, then test it freshly

For the overgeneralised rule, use equal-whole representations to make the condition visible: when positive fractions have the same numerator, the larger denominator gives smaller equal parts, so that fixed number of parts has a smaller value. When the numerators differ, both quantities need attention. Equivalent fractions provide one dependable comparison route.

Then give two new comparisons: three-fifths against three-tenths, followed by three-tenths against one-fifth. Three-fifths is larger than three-tenths. Three-tenths is also larger than one-fifth, because one-fifth equals two-tenths. Ask learners to explain why the first comparison permits the same-numerator shortcut while the second requires a different justification.

A later check can vary the representation again, perhaps using a number line. Do not demand that every learner use the longest method forever. The aim is intelligent choice: a shortcut used within its conditions, with a route back to meaning when those conditions change. The questioning sequence has now done more than mark answers. It has helped locate a boundary, teach it and inspect whether the correction travels.

Worked teaching case: an inference is not a fact hidden in the passage

In English, the teacher may need to distinguish three things: understanding what the text explicitly says, drawing a plausible inference, and explaining how textual evidence supports that inference. A student can succeed at one while needing help with another. An undifferentiated instruction to “add more detail” does not identify the missing connection.

Use this original miniature passage: Mira reached the classroom before anyone else. She laid the group’s model on the table, turned the cracked side towards the wall, and checked the corridor. When her teammates arrived, she said, “I brought it early so we would have time.”

Separate what is stated from what is inferred

Begin with: “What do we know happened?” A learner can identify the early arrival, the model’s cracked side, its placement and the check of the corridor. This establishes the text’s explicit content without yet deciding Mira’s motive. If important details are missed, locate them before asking for a sophisticated interpretation.

Then ask: “What might Mira be concerned about? Use two details to support your suggestion.” One plausible response is that she is concerned about her teammates noticing the damage. Turning the cracked side towards the wall could reduce its visibility, while checking the corridor could suggest concern about someone arriving. The language of possibility matters because the passage does not directly state her thoughts.

A different well-supported interpretation may be acceptable. Mira might be trying to prepare the model before the group arrives. The teacher should ask how the proposed interpretation accounts for the details rather than reject it solely because it is not the teacher’s first idea. Interpretation is constrained by evidence, but it is not always a single-word retrieval task.

Use a follow-up to locate the weak link

If a student says only, “She is nervous,” ask which detail supports that. If the learner quotes the cracked side but does not explain its significance, ask what turning it towards the wall might accomplish. If the learner claims that Mira certainly broke the model, ask where the passage establishes who caused the damage. The point is not to trap the learner; it is to separate an observation, an inference and an unsupported addition.

These responses lead to different teaching. A vocabulary barrier needs clarification. An evidence-selection problem needs comparison between relevant and irrelevant details. An analytical-link problem needs modelling of how an action can support a proposed motive. An overclaim needs a clearer boundary between what is possible, plausible and explicitly established.

A useful final question is: “Name one thing this passage does not allow us to conclude.” Possible answers include who caused the crack, whether the teammates already knew about it, or whether Mira’s explanation was deliberately dishonest. This checks restraint as part of reading, rather than treating confidence or dramatic interpretation as proof of analytical quality.

Check the skill on a new text

After discussion, provide a fresh original passage: Rafi read the message twice, slipped the invitation into his notebook, and told his sister he had no plans for Saturday. Ask for a plausible interpretation, supporting details and a limit on what can be concluded. Do not ask students to reproduce the Mira answer with a different name. The new text calls for the same relationship between evidence, inference and uncertainty.

Evaluate whether the learner now supplies that relationship independently. The student may choose a different interpretation from a peer and still reason well. Conversely, a familiar interpretation expressed fluently may remain unsupported. The questioning job is to make the quality of the connection visible, not to reward the closest imitation of the teacher’s preferred wording.

Worked teaching case: reading data before explaining it

The following numbers are invented for a teaching example. Imagine a stated investigation using equal volumes of warm water in otherwise identical cups in the same room. Cup P is uncovered and Cup Q has a lid. Both start at the same temperature. The table represents one trial; it is not a report of an experiment conducted by eduKate.

Time in minutes	Cup P temperature, °C	Cup Q temperature, °C
0	70	70
5	61	65
10	55	61

Begin with the observation the explanation depends on

Ask: “Which cup has the greater temperature decrease over the first ten minutes, and how large is that decrease?” Cup P decreases by 15°C; Cup Q decreases by 9°C. This question checks reading the starting and ending values and calculating a change. It should not be treated as a complete check of understanding heat transfer.

If a learner answers only that Cup Q is warmer, ask whether the question concerned the final temperature or the amount of change. If the learner selects Cup P but calculates 70 minus 61, inspect which column and time interval were used. If the calculation is correct, ask for a sentence comparing the changes. Each follow-up concerns a different part of the task.

Test whether the claim fits the supplied evidence

Now ask learners to compare two claims: “Cup Q showed a smaller temperature decrease in this trial” and “A lid prevents any cooling.” The first matches the table. The second does not: Cup Q’s temperature still decreased from 70°C to 61°C. A student who selects the second statement may be repeating an overstrong explanation instead of checking it against the numbers.

Ask: “Which observation rules out the stronger claim?” This requires the learner to use evidence to constrain an explanation. It is different from asking for a memorised sentence about lids. Once the data reading is secure, the teacher can connect the result to the heat-transfer ideas already taught, while keeping the distinction between the observed pattern and an explanatory model.

The question “What does this prove about every cup?” should not invite an unlimited generalisation. These illustrative observations concern a particular stated setup and one trial. They support describing that trial; stronger general claims need consideration of repeatability, measurement and other conditions. This boundary can be taught without requiring students to give an advanced research-methods essay.

Make the next question a genuine variation

Ask: “Would looking only at the final temperature be enough if the two cups had started at different temperatures?” The learner should recognise that a final value alone would no longer describe the size of the change. The teacher can provide a new set of invented starting and ending values if calculation practice is needed.

Finally, ask learners to write one supported observation and one question the table cannot answer. The supported observation might compare temperature decreases. An unanswered question might concern whether the same difference would appear in repeated trials. The teacher is checking whether the learner can distinguish data, interpretation and missing information, rather than treating every fluent science sentence as equally warranted.

This sequence also shows why instructional diagnosis should stay specific. A learner who misreads the interval does not necessarily need the scientific mechanism explained again. A learner who reads accurately but claims that no cooling occurred needs to reconcile wording with evidence. A learner who handles both can move to a more demanding explanation or comparison.

Discussion can build understanding, but agreement is not the final check

Discussion should not be dismissed because it changes answers. Changing an answer for a better reason is part of learning. The important question is what changed: the learner’s understanding, the willingness to follow a confident peer, or both. A group’s final answer alone cannot always separate those possibilities.

In a 2009 study, Smith and colleagues examined peer discussion in an undergraduate genetics course using an initial concept question and a subsequent similar question answered individually. Their results supported gains in understanding through discussion, rather than explaining improved answers only as copying a knowledgeable peer. The setting and task matter: this does not establish an identical effect for every age group, subject or discussion format.

A practical classroom sequence is to collect an individual first attempt, let learners compare reasons, and then ask a fresh individual question. During discussion, give the group a reasoning task: identify the point of disagreement, compare evidence, or explain why a tempting alternative fails. “Agree on an answer” alone can encourage the fastest route to agreement rather than the strongest reasoning.

In a three-student group, each learner might first write a reason. One explains a route, another checks it against the task, and the third identifies a question or possible exception. Rotate those responsibilities. No learner should become the permanent answer supplier while others become permanent listeners.

The teacher still needs to resolve consequential errors. A misconception does not become valid because a group agrees with it. After listening, clarify what the evidence or disciplinary rule supports and ask for a new attempt. The wider distinction between shared activity and individual learning is explored in How Group Study Works.

Feedback should give the answer somewhere to go

Questioning loses much of its practical value when the teacher gathers responses but never helps learners use the information. A mark can record an outcome; it does not necessarily specify the next action. The learner needs to know what to reconsider, what to change and what a better attempt would require.

AERO’s feedback recommendations within its formative assessment guide emphasise specificity, relevance to the learning goal and actionable improvement. In the fraction case, that might mean revisiting the condition attached to a comparison rule. In the reading case, it might mean adding the missing explanation between a detail and an inference.

A correction should be followed by work. Ask learners to repair the reasoning, not merely copy the correct answer. Then use another task that requires the repaired relationship. Repeating the teacher’s explanation immediately can be a useful supported step, but it should not be mistaken for a complete independent check.

When the learner is already reasoning well, feedback need not manufacture a weakness. It can name the sound decision and introduce a worthwhile extension: a less familiar context, a competing interpretation or a new constraint. Questioning should make readiness visible as well as difficulty. Otherwise it becomes a system that only notices what students cannot do.

Build a question sequence across the lesson, not a collection of interruptions

A coherent sequence has a beginning, a turning point and a return. At the beginning, ask for the knowledge needed to enter the lesson. During explanation, sample the relationship that the next step depends on. During practice, inspect method selection and execution. Before independent work, check whether the support can reasonably be reduced. At the end, ask for something that informs the next lesson.

Consider a lesson on expanding a simple bracketed expression. An opening question checks what multiplication outside brackets acts on. A worked example then makes the distribution visible. A comparison question presents 3(x + 2) and 3x + 2 and asks whether they are equivalent. Substituting x = 4 gives 18 for the first expression and 14 for the second, exposing the difference.

The teacher can ask where the missing multiplication occurred, then use a fresh expression for practice. The next independent task should not be introduced by telling learners the exact correction it is intended to test. They should have an opportunity to select and execute the appropriate reasoning themselves. If a necessary prompt is supplied, record that as part of the learning sequence.

An exit question should be short enough to answer and specific enough to use. “What did you learn?” may be useful for reflection, but “Explain why these two expressions are not equivalent” gives more direct evidence about the stated objective. The teacher then needs a plan for reading the responses and acting on them, rather than collecting papers that never influence subsequent teaching.

Return to important ideas after time has passed. The IES study guide separately recommends spacing learning and using quizzing to bring key content back into use. In a question sequence, a later check asks a different question from an immediate one: not only whether the learner could follow today’s lesson, but whether the relevant knowledge remains available.

There is no need to turn every return into a formal test. One well-chosen comparison, a short explanation or an unfamiliar application can be enough to decide whether to revisit the topic. A larger decision, such as declaring a broad area mastered, requires broader evidence than a single exit response.

Use small-group visibility without turning differences into fixed labels

A small group offers room to follow individual reasoning closely, but the teacher still needs a plan. It is possible to spend most of a lesson speaking with the most responsive learner while assuming the others are following. It is also possible to over-help every learner until all answers look successful. Neither pattern makes full use of the information a small group can provide.

Use a shared question to locate current differences, then vary the next demand. One learner may need to explain a prerequisite. Another may need to justify a step. A third may be ready to construct a counterexample. The difference should arise from the evidence in front of the teacher, not from a permanent ranking of the students.

Return to a shared task when useful. Learners can compare different representations of the same relationship, or examine why two routes lead to the same result. The group does not need to fragment into unrelated lessons. A common intellectual object can support different levels of responsibility, provided each student has meaningful work to do.

That is the practical connection to How Can Three Students Share a Class Without Receiving the Same Teaching? The aim is not constant individualisation of everything. It is to avoid giving every learner the same next move when their answers have revealed different needs.

A reusable question bank organised by the next decision

The following prompts are starting points for teacher design. Replace the general language with the specific concept, text, representation or task. Do not use the whole bank in every lesson. Select the prompt that can resolve the uncertainty you actually face.

To check access to a prerequisite: “Before we use this idea, what does this term mean here?” Follow with a simple application if a definition alone is insufficient. A fluent definition may still need to be connected to the task.

To check task interpretation: “What are we being asked to find, explain or compare?” Ask the learner to identify the relevant condition. This can separate a problem with reading the demand from a problem with the disciplinary content.

To inspect a first decision: “What would you do first, and what in the question makes that a reasonable start?” The justification helps distinguish deliberate selection from following the method used most recently.

To inspect the meaning of a representation: “What does this line, symbol, label or phrase stand for in the original task?” This is useful when learners can manipulate a representation but may have lost contact with what it represents.

To check a conceptual boundary: “Which feature makes this an example of the idea, and what would need to change for it not to qualify?” Supply a nearby alternative if generating one would create an unnecessary additional demand.

To inspect evidence: “Which part of the supplied information supports that claim?” Then ask whether the evidence supports the full strength of the claim or only a narrower statement.

To inspect a possible error: “Where is the first step that no longer follows from the previous one?” Ask for the governing rule and a repair. Finding the first invalid step is often more informative than merely identifying the wrong final answer.

To test a changed condition: “What stays the same in your reasoning when this feature changes?” The teacher should know which feature is relevant and avoid presenting a supposedly equivalent task that actually requires an additional untaught idea.

To inspect a correction: “What did you change, and why is the new version better?” Ask for a fresh application afterwards when independent use matters. A copied correction is not the same evidence as a reasoned one.

To inspect uncertainty: “Which part can you justify, and which part are you still unsure about?” This gives the learner a more precise response than a global declaration of understanding or confusion.

To inspect checking: “What could you do to discover whether this answer is wrong?” An estimate, substitution, textual comparison or counterexample may be appropriate. The check should fit the subject and the claim.

To hand over learning decisions: “Which next question would help you find out whether you can do this independently?” The learner begins to design evidence rather than wait for the teacher to provide every check. This connects questioning with metacognitive planning, monitoring and evaluation.

Keep a small record of decisions, not a permanent catalogue of student defects

A useful teaching note can be brief: the task, the observed response, the support provided and the next check. For the fraction case, it might read: “Compared same-numerator fractions correctly; overgeneralised denominator rule when numerators changed; used equivalent-fraction example; recheck with fresh pair next lesson.” That is a plan for action rather than a judgement about the learner’s identity.

Include evidence of progress. When a learner can now explain the boundary independently, update the note. An old difficulty should not become the permanent lens through which every later answer is interpreted. A record is useful only if it changes as learning changes.

Use proportionate privacy safeguards. Keep identifiable student work within the appropriate authorised teaching environment, collect only what serves the educational purpose, and avoid turning a learning record into a store of unnecessary personal speculation. A classroom question is not a basis for diagnosing a medical condition, a personality trait or a family circumstance.

For professional sharing, use fictional examples or properly authorised, appropriately anonymised material. Removing a name alone may not remove identifying context. The worked scenes in this guide are invented precisely so the design can be discussed without exposing a real learner’s difficulties.

When to stop questioning and teach

Questioning should not become an ideology in which every piece of knowledge must be extracted from the learner. Some ideas are new. Some prerequisites are missing. Some tasks require a demonstration before a useful independent attempt is possible. The teacher’s responsibility is to provide the help the learning problem calls for.

Stop an unproductive sequence when successive prompts merely narrow the answer until the learner can guess it. At that point, the exchange may be consuming more attention than a clear explanation would. Name the missing relationship, show an appropriate example, and then ask a fresh question that gives the learner a meaningful part to perform.

Also stop when you already have sufficient evidence for the immediate decision. A student who has justified a straightforward step need not defend it through five increasingly artificial questions. Use the available time for worthwhile application. Questioning is a tool for learning and judgement, not a demand that every action be accompanied by a public defence.

In this sense, good questioning includes restraint. The teacher knows when a follow-up will clarify, when it will merely repeat, and when a different teaching mode is required. The question should serve the learner’s progress rather than the teacher’s desire to keep the conversation in one preferred format.

For parents: ask one useful question, then let the answer matter

Parents do not need to turn homework into an oral examination. Start by asking the child to identify the task or explain a decision already made. “What is the question asking?” and “Show me the first place you became unsure” are often more useful starting points than repeatedly asking why the child does not know the answer.

Give room for a response. When help is needed, make the help explicit rather than disguising every explanation as a question. “Let us look at one example together” is clearer than a chain of hints designed to make the child say a preselected word. Afterwards, a small fresh task can show which part the child can now attempt without the same help.

Do not equate frustration with unwillingness or a slow answer with lack of ability. Locate the task-specific difficulty and share concrete evidence with the teacher when needed. An accurate note about where the child stopped is more useful than a broad label. The wider role of a delayed, supported or independent homework attempt is discussed in How Homework Works in Learning.

Using AI to draft questions without outsourcing judgement

AI-generated question drafts should be treated as candidates for review, not as an answer bank that is automatically correct. Before using a generated item, solve it, check its assumptions, examine whether the requested conclusion follows and inspect every alternative. A polished question can still have two defensible answers or require knowledge that has not been taught.

A useful drafting brief states the exact concept, the learner’s relevant prior knowledge, the misconception or boundary to examine, the permitted response format and the intended teaching decision. Ask for a proposed answer and a rationale for each alternative, then verify them. Do not accept a claim that a particular wrong option uniquely identifies a student’s mental state.

Keep identifiable learner material out of an external tool unless its use is authorised and appropriate for that environment. There is usually no need to supply a student’s name or private history to design a fraction comparison or an original reading passage. Fictional task-specific descriptions are often sufficient for the drafting job.

When learners use AI for practice, preserve opportunities to answer before seeing a generated explanation and to attempt a fresh task afterwards. Do not infer independent understanding from an answer assembled with external reasoning. The educational question remains what the learner can now do, as discussed in How AI-Assisted Study Works.

A practical review of one questioning episode

After a lesson, choose one consequential exchange rather than trying to review every question. Reconstruct the learning objective, the wording, the response conditions, the answer and the next action. Then ask whether the information actually justified the decision made.

Did the question examine the intended concept, or did it mostly test vocabulary, speed or familiarity with the format? Did each relevant learner have a chance to respond? Did a hint supply the decision you later credited to the student? Did you accept a correct choice without inspecting a questionable reason? Did a wrong answer lead to teaching that addressed its likely cause?

Look for successful design as well. A concise contrast may have clarified a boundary. A written response may have revealed reasoning that oral participation missed. A fresh question after discussion may have shown that a learner could now explain independently. Keep those question-and-response patterns for reuse, while remaining ready to adapt them to a different class.

The useful improvement may be small: changing one ambiguous phrase, collecting answers before discussion, allowing more preparation, or adding a follow-up that distinguishes two plausible explanations. Better questioning does not always require a new teaching philosophy. It can begin with a more accurate relationship between one question, one response and one decision.

Evidence, interpretation and the limits of this guide

This article combines established educational guidance with original teaching designs. The external sources support particular principles: using assessment information to adjust instruction, asking for explanations, considering thinking time, offering appropriate response modes, and checking what learners can do after discussion. They do not validate this entire article as a single tested programme.

The fraction sequence, the Mira and Rafi passages, the cooling-data table and the reusable prompts are editorial examples. Their calculations and internal logic can be checked, but their classroom effectiveness would still depend on learners, instruction, context and implementation. No attainment gain, diagnostic accuracy rate or universal waiting time is claimed.

Sources were reviewed for this edition on 5 September 2026. The cited research is not all recent; older studies are identified by their dates rather than presented as new findings. Official guidance, classroom demonstrations and primary-study abstracts also provide different kinds of evidence and should not be treated as interchangeable.

Sources and further reading

Institute of Education Sciences / What Works Clearinghouse, 2007: Organizing Instruction and Study to Improve Student Learning. The official recommendation summary includes explanatory questioning, quizzing and spaced learning. Its separate recommendations should not be collapsed into a guarantee for any particular lesson script.

Australian Education Research Organisation: Monitor Progress, first published in 2024, with a displayed update of 14 May 2026; and Formative Assessment: Practice Guide, first published in 2021, with a displayed update of 7 July 2026. These are practice guidance, not evaluations of the original examples in this article.

AERO classroom illustrations: Monitor Progress: Primary and Secondary. The published video transcript illustrates participation routines and teacher responses. Individual practitioners’ routines or thresholds should not be mistaken for universal requirements.

Mary Budd Rowe, 1972: Wait-Time and Rewards as Instructional Variables: Their Influence on Language, Logic, and Fate Control. The ERIC record supplies the report’s abstract and historical study context.

CAST, 2024: UDL Guidelines 3.0, consideration 5.1: multiple media for communication. This supplies an accessibility-design principle, not a claim that every response medium assesses every learning goal equally.

Smith and colleagues, 2009: Why Peer Discussion Improves Student Performance on In-Class Concept Questions, Science, 323, 122–124, DOI 10.1126/science.1165919. The discussion here is bounded to the study’s reported design and conclusion in its primary abstract, not a fresh analysis of its underlying data.

The question is finished when it has helped learning move

Return to the three students with the same correct answer. A teacher could have accepted the agreement and continued. Instead, a carefully chosen follow-up distinguished a sound comparison, an overextended rule and an answer that had not yet been independently produced. The teacher now had something more useful than a pleasing chorus of correctness: a reason to choose different next moves.

That is the discipline of questioning. Decide what matters. Create a fair opportunity to respond. Inspect the answer without pretending to read the learner’s mind. Teach where the evidence calls for teaching, extend where it shows readiness, and return with a question that tests what has become available to the learner.

A strong question does not merely end in an answer. It changes what the teacher can responsibly do next.

Return to the series map in How Teaching Works, revisit the modelling stage in How Worked Examples Work in Teaching, or hand the question-design role to the learner through How Question Generation Works in Learning.

Teaching route: return to The Tutor Handbook for tutoring decisions, or use How Learning Works for the wider learning mechanism map.

Questioning that makes learner thinking visible

A useful teaching question is not merely difficult or open-ended. It is designed so the learner’s answer helps the tutor distinguish between plausible learning states and decide what to do next. The question is finished only when the answer changes—or confirms—the teaching route.

Use How to Improve Questioning for the broader learner/parent skill of asking better questions. This page owns questioning as a teaching mechanism. The Tutor Handbook’s Wait-Time and Prompt-Latency Gate owns the professional decision about when to stay silent, rephrase or hint.

Subject	Question that reveals thinking	What the answer can separate
English	“Which words in the passage support that inference, and which part is your interpretation?”	Passage understanding vs evidence selection vs unsupported inference.
Mathematics	“Before calculating, what relationship do you think this problem is describing?”	Arithmetic weakness vs representation/method-selection weakness.
Science	“What changed first, and how does that change produce the observation?”	Keyword recall vs causal mechanism.
Additional Mathematics	“What structure in this expression makes that identity or transformation useful?”	Procedure recall vs structural recognition and selection.

Wait long enough to preserve evidence

A tutor who rephrases too quickly can accidentally answer the question. A short silence may reveal retrieval, planning or uncertainty that the tutor needs to see. But silence should not become ritual: when the learner is genuinely Blocked, a narrower prompt may be the right teaching move. The technical wait-time gate should guide that decision.

The follow-up should test the learner’s model

When a learner gives a correct answer, change one feature and ask again. When the answer is wrong, choose a follow-up that separates possible causes rather than simply pointing toward the correct response. This makes questioning diagnostic without turning the lesson into continuous testing.

Question architecture: decide what kind of evidence you need before asking

A useful teaching question begins with the decision the tutor may need to make next. If the learner answers correctly, what will that allow the tutor to infer? If the learner answers incorrectly, what different explanations are possible? A question that cannot improve the next teaching move may still create discussion, but it is weak diagnostic evidence.

Question architecture means designing the prompt, the response conditions, the wait time, the follow-up and the fresh check as one evidence sequence. The wording matters, but the surrounding conditions matter just as much.

Question job	Example	What it can reveal	Common mistake
Retrieve	“Without notes, what is the formula/definition/relationship?”	Availability of stored knowledge.	Giving cues that turn recall into recognition.
Interpret	“What is this question asking you to find?”	Task understanding and scope.	Rephrasing so strongly that the tutor supplies the interpretation.
Represent	“How could you show this relationship?”	Model/diagram/equation selection.	Offering the representation before learner choice is visible.
Explain mechanism	“What changes first, and how does that cause the observation?”	Causal structure, not keyword recall.	Accepting correct terminology without the relationship.
Justify	“Why does that step follow?”	Reasoning behind a correct answer.	Turning every correct answer into an exhausting oral defence.
Discriminate	“Which of these two cases uses the same rule, and why?”	Boundary between similar methods/concepts.	Choosing examples that differ in too many ways.
Transfer	“What changes if the context/representation changes?”	Flexible use beyond familiar format.	Changing both target and context so failure is uninterpretable.
Monitor	“Where are you least certain in this solution?”	Self-monitoring and help-seeking.	Using confidence as a substitute for correctness.

Question sequences should narrow uncertainty, not just increase difficulty

A strong sequence is not simply easy question → harder question → hardest question. It should reduce uncertainty about the learner model. The first question may establish what is known. The second changes one feature to test the rule. The third removes a support or asks for transfer. Each response changes what the tutor needs to ask next.

A five-question diagnostic sequence

1. What do you notice? Gather the learner’s spontaneous representation or interpretation.
2. What do you think the task is asking? Separate task reading from subject execution.
3. What is your first move, and why? Expose method selection.
4. What would make this move invalid? Test boundary awareness.
5. Can you do a fresh version with one feature changed? Test transfer rather than memory of the conversation.

The tutor does not need to ask all five every time. The sequence shows how questions can progressively locate a weak link instead of circling the same uncertainty.

English questioning: reveal the reading and writing decision

Comprehension

In comprehension, a vague “why?” can hide several jobs. Ask the question that matches the suspected boundary: “Which words in the passage support that?”, “Is that stated or inferred?”, “What part of the question limits your answer?”, or “What would make this inference too broad?”

If the learner gives a correct answer but cannot point to evidence, do not immediately conclude the answer was guessed. A follow-up can distinguish memory, background knowledge, inference and evidence selection. Then use a fresh passage because the explanation may have taught the original item.

Writing

Questioning in writing should preserve authorship. “What are you trying to make the reader understand here?” reveals purpose. “Which sentence carries the claim?” reveals structure. “What evidence does this paragraph still need?” reveals development. Avoid questions that smuggle the tutor’s preferred sentence into the learner’s draft.

A good writing follow-up returns the decision: “You have two possible examples. Which one proves your claim more directly?” The learner must choose and justify rather than copy a suggested revision.

Grammar and editing

Instead of “What is the tense?”, ask what time relationship the sentence needs to express. Instead of pointing at the error, ask where the sentence stops matching its intended meaning. This shifts grammar from label recall to meaning control.

Oral

Questioning should not reward the fastest speaker. Use deliberate wait time, then ask one extension question tied to the learner’s answer. A stronger oral sequence is answer → reason → example → boundary, not a rapid list of unrelated prompts.

Mathematics questioning: expose representation and method selection

Primary problem solving

A child can calculate accurately and still fail problem sums because representation is weak. Questions such as “What quantity is the whole?”, “What stays the same?”, “What changes?”, and “Can you show the relationship without calculating yet?” reveal the structure before arithmetic hides it.

If the tutor asks “Is this ratio?” too early, the topic label becomes a hint. When method selection is the target, questions should preserve that decision.

Algebra

A correct algebra line can come from memorised shorthand. Ask “What operation keeps both sides equal?” or “What must stay true after this step?” to reveal whether equality is understood. If the learner can explain the invariant, later compressed notation may be safe. If not, the shorthand is carrying understanding.

Geometry

Instead of telling the learner which theorem to use, ask what information the diagram guarantees and what remains unproven. Then ask which relationship would connect the known facts to the target. This keeps theorem selection with the learner.

Graphs and data

Ask in layers: What does the axis measure? What pattern is visible? What numerical evidence supports that description? What can we not conclude? The final boundary question prevents the learner from turning correlation or a limited range into a universal claim.

Additional Mathematics questioning: keep abstraction with the learner

A-Math questions often fail when the tutor announces the method. “Use differentiation” may produce a correct solution while leaving method selection untested. A stronger question asks what quantity is changing, what information the derivative would provide, or what representation makes the relationship easier to analyse.

For trigonometric equations, ask what form would make the equation solvable rather than naming the identity. For functions, ask what input/output relationship the notation expresses before manipulating symbols.

Science questioning: distinguish fact, mechanism and evidence

Facts

A direct retrieval question can establish whether the learner knows the term or fact. Do not overinterpret a correct keyword as understanding of the mechanism.

Mechanisms

Ask causal questions one link at a time: “What changes first?”, “What does that change affect?”, “How does that produce the observation?” If the learner cannot answer the first link, giving the final explanation hides the first weak link.

Experimental design

Questions should reveal whether the learner understands what evidence would answer the question: “What are you changing?”, “What are you measuring?”, “Which alternative cause needs to be controlled?”, “How would another student repeat this?”

Avoid turning the questioning sequence into a checklist whose wording gives away the variables. As competence grows, remove the labels and ask the learner to design the evidence route.

Data interpretation

Ask the learner to separate observation from explanation: “What does the graph show?”, then “What might explain it?”, then “What extra evidence would increase confidence?” This prevents explanation from overwriting the actual data.

Questioning in a three-student tutorial

Three students create a special opportunity: one answer can become evidence for the others without becoming their answer. The tutor must protect private thinking before public discussion.

Think first, then share

Give all three students a short private attempt or prediction before anyone speaks. This stops the fastest learner from becoming the group’s method selector. After sharing, ask another learner to agree, disagree or add a reason—not merely repeat.

Use sideways follow-ups

A sideways follow-up sends the idea to another learner: “Ciara says the whole changes here. Denise, what in the question supports that?” This keeps the group engaged while preserving individual reasoning.

Protect the original learner

Do not turn one learner into a permanent explainer or use their error as the group’s public identity. Questions should move the thinking around the room, not assign fixed roles.

Wait time is part of the question

A question followed immediately by a rephrase, hint or tutor answer is not the same question as one that gives the learner time to formulate a response. Wait time changes who carries the thinking.

The Tutor Handbook’s Wait-Time and Prompt-Latency Gate owns the technical judgement about when silence is productive. The public rule is simpler: wait long enough to see whether the learner is thinking, but not so long that the learner has no route and the silence becomes unproductive stuckness.

What to watch during wait time

• Does the learner reread or inspect the task?
• Do they begin a representation or tentative step?
• Do they retrieve quietly before speaking?
• Do they look toward the tutor for a cue rather than the task?
• Do they repeat the same failed action without new information?
• Can they state what is blocking the next move?

Follow-up questions should not contain the missing answer

A partly correct response tempts tutors to ask leading questions. “So if surface area increases, evaporation becomes…?” nearly supplies the missing conclusion. A better follow-up asks which relationship connects the changed surface area to the observed result.

The Tutor Handbook’s Follow-Up Question gate owns the professional boundary. The acceptance test is whether the learner later supplies the missing reasoning on a fresh task without the tutor’s funnel.

Return questions instead of answering every help request

Learner questions are valuable evidence. “Which formula do I use?” may reveal missing method selection, not missing formula knowledge. The tutor can answer, clarify or return part of the question depending on what the learner still needs to practise.

A bounded return sounds like: “You know three formulas here. What information in the question would distinguish them?” It preserves help-seeking while handing the decision back.

Questioning by learner state

A question that is useful for a Transfer-Ready learner may be harmful for a Blocked learner. Questioning should therefore change with the capability state rather than follow a fixed script.

Blocked

The learner cannot form a usable first move. Questions should reduce uncertainty enough to reopen thinking without hiding the missing prerequisite. “What do you already know about these quantities?” may help. A broad “What do you think?” can leave the learner stranded.

If no question produces useful movement, stop questioning and teach. Interrogating a missing concept does not create the concept.

Fragile

The learner can perform with recent support but the capability is unstable. Questions should encourage retrieval and self-explanation: “What did you use last time?”, “Why does that relationship hold?”, “Can you do the first step without the example?” Avoid immediately re-teaching after hesitation; allow enough wait time for retrieval.

Stable

The learner succeeds on familiar work. Questions should test boundaries and method selection: “How do you know this is the right method?”, “What would make it the wrong method?”, “Can you represent it another way?”

Transfer-Ready

Questions should vary context, representation and assumptions. “Would your explanation still hold if this condition changed?” “What is structurally the same about these two problems?” “Which part of your rule transfers and which does not?”

Examination-Ready

Questioning becomes lighter during authentic performance. Before or after the task, ask about pacing, selection, checking and recovery. Do not turn the tutor’s question sequence into a hidden examination scaffold.

State	Question job	Avoid
Blocked	Locate first weak link or prerequisite.	Long chains of leading questions that manufacture an answer.
Fragile	Retrieve, explain and reduce support.	Immediate hints that keep capability permanently warm.
Stable	Test boundaries and method selection.	More questions that merely repeat the familiar routine.
Transfer-Ready	Change conditions and test flexible use.	Novelty that quietly changes the target.
Examination-Ready	Reflect before/after authentic task.	Question-by-question coaching during the performance evidence.

Questions for misconceptions

A misconception is not merely a wrong answer. It is a wrong model that can generate several wrong answers consistently. Questioning should expose what the learner expects to happen if their model is true.

Prediction before correction

Ask the learner to predict the result in a new case. Then compare the prediction with evidence or a counterexample. This makes the model visible and creates a reason to revise it.

Contrast cases

Put two cases side by side that differ only in the decisive feature. Ask which difference matters and why. Contrast is especially useful for algebraic cancellation, grammar agreement, Science fair tests and inference boundaries.

Do not debate the misconception into submission

Once the wrong model is clear, teach the better model. Endless questioning can become adversarial and may simply train the learner to guess what the tutor wants.

Questions for uncertainty

Uncertainty is not the same as ignorance. A learner may know the method but be unsure about one assumption. Good questions help locate the uncertainty without turning confidence into correctness.

• “Which step are you most certain about?”
• “Where does your confidence drop?”
• “What information would you need to decide between these two methods?”
• “What evidence would make you change your answer?”
• “What part of your answer is observation, and what part is inference?”

These questions train calibrated judgement. The learner becomes better at asking for precise help and at deciding when checking is necessary.

Questions for self-checking

“Check your work” is too vague for many learners. A better questioning sequence identifies what type of error is plausible: “Which copied values matter?”, “Can you reverse the operation?”, “Does the answer fit the size of the original quantities?”, “What sentence in the passage proves your claim?”

As self-checking improves, the questions should become self-prompts. The tutor’s external question is temporary support.

Questions for planning

Planning questions should reveal priorities without writing the plan for the learner. “What must be decided before you can start?”, “Which part depends on another?”, “What can wait?”, “What would a complete answer need?”

If the tutor supplies the sequence every time, planning remains adult-owned. Fade toward one broad prompt, then independent planning.

Questions for retrieval

Retrieval questions should require the learner to produce from memory before cues appear. Recognition questions can still be useful, but they are weaker evidence of recall. If the learner fails, reduce the retrieval demand deliberately rather than accidentally through increasingly obvious hints.

Questions for transfer

A transfer question changes something that should not matter to the underlying rule. The learner must recognise the structure despite the changed surface. Useful changes include wording, representation, context, order, scale or neighbouring topics.

Do not call a question “transfer” just because it is difficult. Transfer depends on structural continuity beneath surface change.

Questioning and accessibility

A learner may need more processing time, clearer wording, larger text or another legitimate access condition. Those supports should not be confused with giving away the reasoning. Use the Accessible Learning Tasks owner to decide whether a rephrase preserves the target or silently changes it.

For example, breaking a long procedural instruction into bullet points may preserve a Mathematics reasoning target. Rewriting the mathematical relationship in simpler conceptual terms may supply the representation that was being assessed.

Questioning and language load

A subject question can fail because the learner does not understand the wording. To diagnose, ask the learner to paraphrase the question. If the paraphrase is wrong, the issue may be language/task interpretation rather than subject execution. Simplify only enough to locate the boundary, then route the language need appropriately.

Questioning and cognitive load

Long multi-part questions can overload working memory. Ask one coherent question at a time when the goal is to observe a specific reasoning step. Multiple prompts in one sentence can create apparent confusion that belongs to the question design, not the learner.

Questioning and speed

Fast answers are not necessarily stronger answers. Slow answers are not necessarily weak. Response time should be interpreted relative to task purpose, learner state and access conditions. The Wait-Time gate exists because visible speed can distort tutor judgement.

Questioning and confidence

Ask learners to estimate confidence only when the estimate changes the decision. Confidence can help identify calibration problems, but a confident wrong answer still needs conceptual repair and an uncertain correct answer may need evidence-based self-trust.

Questioning and motivation

Do not use questioning as disguised moral pressure: “Why didn’t you study?” rarely improves the learning model. Ask about the conditions: “When did you try?”, “Where did you stop?”, “What made the first step hard?”, “What did you do after getting stuck?”

Behaviour becomes more interpretable when the questions produce task evidence rather than blame.

Questioning and error analysis

After an error, ask about the route before the final answer. “What did you think this symbol meant?”, “What relationship were you using?”, “What made you choose this method?” The goal is to locate the earliest unreliable step.

If the learner cannot reconstruct their own reasoning, the tutor may need a fresh similar item rather than interrogating memory of the original attempt.

Questioning and feedback

A question can turn feedback into learner action: “What will you change first?”, “How will you know the revision worked?”, “Can you use the comment on a fresh example?” This is stronger than asking whether the learner “understands the feedback”.

Questioning and independent performance

Once the tutor has asked enough questions to teach the missing idea, the next evidence must eventually become quieter. Independent performance cannot be measured through an endless scaffold of tutor questions. Ask, teach, fade, then observe.

Question quality audit

Audit question	Strong sign	Warning sign
Does the question have one job?	Tutor can name the evidence sought.	Question tries to diagnose, hint and teach at once.
Does wording preserve the target?	Learner still owns the decisive reasoning.	Question contains the missing answer.
Is wait time appropriate?	Learner has real opportunity to think.	Tutor rephrases instantly or waits without any route.
Can the answer change the next move?	Tutor can branch based on response.	Answer is collected but ignored.
Is the question necessary?	It reveals something not already known.	Tutor asks from habit or to fill silence.
Is there a fresh check?	Teaching generated by questioning is retested elsewhere.	Original corrected answer is treated as proof.

The question ladder

When a learner is stuck, questions can move from least to more informative support. The ladder should be used flexibly, not mechanically.

1. “What do you notice?”
2. “What is the task asking?”
3. “What do you already know that might matter?”
4. “What representation could make the relationship visible?”
5. “Which two options are you deciding between?”
6. “What would be a useful first step?”
7. “Would a related example help?”
8. Teach/model the missing element when questioning no longer has educational value.

Stop questioning and teach

Questioning is not superior to explanation. If the learner lacks a prerequisite, has no usable model or repeatedly guesses despite clear wait time, direct teaching is often the better next move. The tutor should not force discovery of knowledge the learner has never had a fair chance to learn.

After teaching, return to questioning on a fresh item to see whether the learner can now use the idea.

Stop teaching and ask

The opposite boundary also matters. A tutor can explain long after the learner is ready to try. Once the learner has enough model to act, ask for a prediction, first move or fresh example. The question transfers responsibility back.

Questioning as a handover mechanism

The deepest role of questioning is not to keep the tutor involved; it is to externalise questions the learner can later ask themselves. “What is the whole?”, “What evidence supports this?”, “What changed first?”, “What would make this method invalid?” become self-prompts.

Successful questioning therefore disappears into learner self-regulation. The tutor does not need to ask forever.

Questioning casebook across the school journey

The questioning job changes as students acquire more knowledge, more independent control and more complex subject demands. The same tutor sentence—“Why?”—can be useful at one stage and empty at another. The cases below show how to choose questions that expose the next useful piece of evidence without turning questioning into a script.

Primary 1 reading: sound, word and meaning

A Primary 1 learner pauses at an unfamiliar word. A weak question is “What word is that?” repeated louder. A stronger sequence first identifies what is available: “What sound does this letter make here?”, “Can you blend the first two sounds?”, then after decoding, “Does that word make sense in the sentence?” The first questions scaffold decoding; the last reconnects decoding with meaning.

If the learner can decode only after every sound is prompted, the tutor has teaching evidence, not independent reading evidence. A fresh word later should carry less prompting.

Primary 1 Mathematics: number bonds

A learner knows 7 + 3 = 10 but does not see why 7 + 5 can become 10 + 2. Ask “How much does 7 need to make 10?” then “Where could that amount come from?” The question sequence exposes decomposition. If the tutor instead asks “Can you split 5 into 3 and 2?”, the decisive decomposition has already been supplied.

Primary 2 English: sentence meaning

A learner selects the wrong pronoun. Rather than “Which noun does it refer to?”, ask the learner to read both possible antecedents into the sentence and decide which meaning remains coherent. The question turns grammar into meaning checking rather than label matching.

Primary 2 Mathematics: problem representation

The learner can calculate but chooses the wrong operation in a story problem. Ask “What is happening to the quantity?” before asking for an operation. Follow with “Can you show the relationship with objects, a bar or an equation?” The representation should come from the learner where possible.

Primary 3 Science: observation versus inference

A learner writes “the plant is unhealthy”. Ask “Could a camera record that exact statement?” This question reveals the observation/inference boundary without giving the observable detail. Then ask “What can you actually see?” The learner may revise to “three leaves are yellow and drooping.”

Primary 3 Mathematics: multiplication meaning

If a learner knows multiplication facts but struggles with grouping, ask “How many groups are there?” and “How many in each group?” Then vary the representation. The goal is to test whether the equation maps onto the structure, not whether the student can recite a fact.

Primary 4 comprehension: inference

The learner answers a character-motive question using a plausible but unsupported idea. Ask “Which action or phrase makes you think that?” If the learner cannot identify evidence, the tutor has located the gap between inference generation and evidence support. On the next passage, the learner should mark the evidence before answering.

Primary 4 Science: fair tests

The learner changes two variables. Ask “If your result changes, which change would you blame?” This can make the confounding visible more powerfully than reciting “change one variable only.” Then ask the learner to redesign the comparison.

Primary 5 Mathematics: percentage and the reference whole

A learner calculates 20% correctly but applies it to the wrong quantity. Ask “Twenty percent of what?” then “Can the whole change during this question?” Those questions target the reference quantity. A later fresh problem should test whether the learner asks those questions internally.

Primary 5 English: composition development

A paragraph contains a claim and one example but little explanation. Ask “What does this example make the reader understand about your point?” The answer should become the missing explanatory bridge. Avoid supplying a sentence starter that contains the conclusion unless modelling is the intended job.

Primary 5 Science: mechanism

The learner uses the correct keyword but skips the causal chain. Ask “What changed first?” and “What did that change affect?” If the learner can supply the chain, the vocabulary was not the main weakness. If not, concept repair is needed.

Primary 6 PSLE preparation: answer scope

A learner writes too much for a comprehension question. Ask “What exactly is the question asking you to prove?” and “Which part of your answer goes beyond that job?” This makes scope visible. In a timed paper, the tutor should not ask these questions live; the learner needs self-prompting.

Primary 6 Mathematics: full-paper checking

After a paper, ask “Which answers did you have a concrete reason to doubt?” and “What verification method would fit each?” This is stronger than “Did you check?” because it reveals whether checking is targeted or ritual.

Secondary 1 English: transition to more abstract comprehension

A learner can locate evidence but writes literal paraphrases for inference questions. Ask “What does the writer expect you to understand without saying directly?” Then ask for the textual clue. This sequence distinguishes inference from evidence support.

Secondary 1 Mathematics: algebraic expressions

The learner writes 3 + x for “three times x”. Ask “If x were 4, what would the phrase mean numerically?” Substitution can expose the language error without naming multiplication. Then return to symbolic form.

Secondary 1 Science: explanation depth

A learner states a fact but does not link it to the observation. Ask “How does that fact produce what we can see in this setup?” This shifts the response from recall to mechanism.

Secondary 2 English: evaluation of evidence

A learner finds a relevant quote but treats relevance as sufficient. Ask “What exactly does this evidence prove?” and “What does it not prove?” The boundary question develops precision and prevents overclaiming.

Secondary 2 Mathematics: algebraic factorisation

The learner factorises only when the chapter heading tells them to. In a mixed question, ask “What structure do you see?” rather than “Should you factorise?” If the learner identifies a common factor or quadratic pattern independently, method selection is becoming visible.

Secondary 2 Science: experimental conclusion

A learner concludes that one variable “causes” the result from a weak design. Ask “What other difference could explain this pattern?” The question reveals whether control variables and causal inference are understood.

Secondary 3 English: discursive argument

The learner has a strong opinion but weak reasoning. Ask “What would a reasonable person who disagrees say?” followed by “Which part of your argument survives that objection?” Counterargument questions expose assumption quality without writing the response for the learner.

Secondary 3 Mathematics: trigonometry

The learner remembers formulas but chooses poorly. Ask “Which sides are known relative to the angle?” before naming a ratio. Later, remove even that question and ask the learner to explain their selection independently.

Secondary 3 Additional Mathematics: functions

A learner manipulates notation without understanding. Ask “What input goes into this function?” “What output comes out?” “What does the composition tell you to do first?” These questions reconnect symbols with operations.

Secondary 3 Science: data and model limits

Ask “Is this result consistent with the model?” rather than “Does this prove the model?” Then ask what evidence would challenge the model. The language of consistency and falsification keeps claims proportional.

Secondary 4 English: final-year writing

During revision, ask fewer questions but make them higher leverage: “What is the controlling claim?”, “Which paragraph contributes least?”, “Where is the first unsupported inference?”, “What would the marker still need to know?” The learner should increasingly run these checks without the tutor.

Secondary 4 Mathematics: paper recovery

After a timed section, ask “Where did the paper state change from productive work to being trapped?” and “What observable cue should trigger leaving next time?” This is a post-performance question. Asking it during the paper would become a scaffold.

Secondary 4 Additional Mathematics: calculus

The learner can differentiate but misreads the interpretation. Ask “What does this derivative represent in the context?” followed by “What would its sign tell you?” The question reconnects technique with meaning.

Question response types need different interpretations

Response pattern	Possible meaning	Useful next move
Correct and immediate	Could indicate strong retrieval, familiarity or cue recognition.	Change representation or ask for reason only if it changes the decision.
Correct after long pause	May reflect deliberate retrieval or fragile access.	Give another item after delay before labelling speed weakness.
Incorrect but reasoned	Model is visible and can be repaired.	Target the earliest wrong assumption.
Correct with weak reason	Could be guess, memorised rule or expression difficulty.	Ask one discriminating follow-up or changed case.
“I don’t know” immediately	Could be blocked, avoidance, uncertainty or habit.	Ask what is known and provide wait time before hinting.
Changes answer after tutor facial cue	Confirmation dependence may be present.	Neutralise tutor response and teach self-verification.
Repeats peer answer	Group agreement may hide individual understanding.	Ask learner to reconstruct privately or apply to fresh case.

Do not confuse more talk with more thinking

A lively lesson can still produce shallow evidence. Learners may speak often because the tutor asks many low-demand questions, because one learner echoes another, or because answers are obvious from the tutor’s phrasing. The Tutor Handbook Learner-Talk Interpretation gate exists to prevent participation from becoming a proxy for learning.

A quieter answer that contains a real model, uncertainty or justification can be more informative than many quick responses. Questioning quality should be judged by the thinking it reveals and the next teaching move it enables.

Participation architecture in a three-student class

• Protect first attempts. Give all three learners time to think before public answers.
• Vary who goes first. Do not let speed define expertise.
• Use write–think–share. A short written response preserves individual evidence.
• Ask for comparison. “What is different about your route?” creates useful peer visibility.
• Return to each learner. Shared discussion should end with individual reconstruction or fresh work.
• Track invisible learners. Silence can hide mastery or confusion; sample both deliberately.

Questions that accidentally give away the answer

Leading form	Why it gives away too much	Better form
“Shouldn’t you multiply here?”	Names the operation.	“What relationship connects these quantities?”
“Because heat increases evaporation, what happens next?”	Supplies the mechanism.	“What effect does the temperature change have on the process?”
“Which quote proves she is angry?”	Supplies the inference.	“What emotion do you infer, and what text supports it?”
“Can you factorise first?”	Names method selection.	“What structure in the expression might simplify the problem?”
“Isn’t this a control variable?”	Contains expected classification.	“What role does this variable play in the comparison?”
“Don’t you need a counterargument?”	Supplies essay architecture.	“What could a reasonable reader challenge in your claim?”

Leading questions are not forbidden. They can be deliberate scaffolds during teaching. The problem occurs when the tutor uses them and later interprets the learner’s answer as independent evidence.

Questioning when the learner is wrong but productive

Do not interrupt every wrong turn immediately. If the route is safe and informative, allow enough working to reveal the learner’s model. A premature correction can erase the evidence that would have made the repair precise.

Intervene when the error is likely to be rehearsed repeatedly, when the learner has no productive route, or when continuing would waste substantial time. The decision is about learning value, not about tolerating error for its own sake.

Questioning when the learner is correct but fragile

A correct answer can still deserve a question if the tutor needs to know whether the route is robust. Ask for a changed case, a reason, or a boundary—not a generic “are you sure?” which can reduce confidence without revealing anything.

Questioning when the learner is correct and strong

Sometimes the best next move is no follow-up at all. If the evidence is already sufficient, excessive questioning wastes time and can turn competence into performance theatre. Move to a new challenge, transfer task or another learner.

Questioning after support

If a learner answers after a hint, record that support mentally or explicitly. The next question should not simply celebrate the correct response; it should help decide whether the hint can fade. A fresh item is often stronger than another verbal probe on the same item.

Questioning after feedback

A question such as “What will you change on the next attempt?” turns feedback into a learner decision. Later, “Did that change survive on the fresh task?” connects questioning to evidence rather than conversation alone.

The questioning evidence contract

A teaching question should have an evidence contract before it has elegant wording. The tutor should know what kind of learner evidence the question is supposed to produce, what support is allowed, how long the learner will have to think, and what the answer could legitimately change in the lesson. Without that contract, questions easily become performance theatre: everyone talks, the tutor feels interactive, but the next teaching move is no clearer.

A useful questioning contract can be stated in one sentence: “I am asking this question to find out whether the learner can ______ under ______ conditions, so that I can decide whether to ______.” The blanks force the tutor to identify capability, conditions and decision before improvising the prompt.

Question job	Evidence sought	Possible next move
Check retrieval	Can the learner bring the idea back without the model visible?	Continue, retrieve again later, or repair if access is missing.
Check interpretation	Does the learner understand what the task is asking?	Clarify language or move to the academic method.
Reveal representation	How is the learner organising the problem or idea?	Teach or compare representations before calculation/writing.
Reveal method selection	Can the learner choose a route when the method is not named?	Vary examples or explicitly teach selection cues.
Reveal mechanism	Can the learner explain why the result follows?	Repair causal chain rather than keyword recall.
Reveal evidence boundary	Can the learner distinguish what the data shows from what it does not show?	Teach claim calibration and retest on new data.
Reveal checking	Can the learner detect and verify an error independently?	Teach verification method or fade adult checking.
Reveal transfer	Can the learner carry the capability into a changed context?	Increase variation or return to prerequisite repair.

Question architecture: ask for one discriminating piece of evidence at a time

A strong diagnostic question separates plausible explanations. A weak question often allows several interpretations. “Do you understand?” is weak because yes/no tells us little. “Which quantity is the whole in this problem, and what in the wording tells you that?” is stronger because it distinguishes reading, representation and method-selection possibilities.

The best diagnostic question is often not the hardest question. It is the question that most efficiently reduces uncertainty. In a three-student tutorial, the tutor should prefer a short discriminating question over a long performance task when the teaching decision only requires one boundary to be checked.

The minimum-discrimination principle

Ask the smallest question that can change the next move. If the uncertainty is whether the learner recognises the denominator relationship, do not assign a six-step word problem. If the uncertainty is whether the learner can control comprehension answer scope, ask for one evidence-bounded response. Large tasks are useful later, once the smaller uncertainty has been resolved.

Question sequences should narrow uncertainty

One question rarely provides the whole diagnosis. A sequence should narrow possibilities without feeding the route. The tutor begins broad enough to let the learner reveal a model, then follows with a question that distinguishes between competing explanations, then decides whether to teach, vary or retest.

1. Ask for the learner’s first interpretation or move.
2. Ask for the basis: “What made you choose that?”
3. Change one decisive feature while preserving the underlying target.
4. Ask the learner to compare the two cases.
5. Teach only the missing relationship that the evidence now supports.
6. Retest on a fresh task without the original question sequence doing the thinking.

The sequence is complete only when the learner later performs without needing the diagnostic conversation. Questioning can reveal and teach, but fresh independent work is the acceptance evidence.

Wait time is part of the evidence

The time between question and tutor intervention changes what the learner gets to demonstrate. A learner who is interrupted after one second may appear unable to retrieve an answer that would have emerged after six. A learner who is routinely rescued after a predictable pause may learn to wait for the prompt rather than continue thinking.

The technical professional boundary remains in the Tutor Handbook Wait-Time and Prompt-Latency Gate. The public teaching rule is simpler: preserve enough silence for the learner to think, watch what happens during that silence, and intervene when the pause stops producing useful evidence.

What productive wait can look like

• The learner rereads the question or points to relevant information.
• The learner writes a partial representation or tries a first case.
• The learner revises an earlier statement without tutor prompting.
• The learner asks a more precise question.
• The learner visibly compares alternatives rather than merely waiting.
• The learner uses an agreed self-check or help-seeking routine.

What unproductive waiting can look like

• The learner repeatedly scans without identifying any relevant feature.
• The learner watches the tutor for a cue instead of the task.
• The learner begins guessing random methods.
• The learner disengages or switches to unrelated activity.
• The same blocked point persists despite sufficient opportunity and no new strategy appears.

The distinction is observational, not moral. Productive silence should be allowed to continue; unproductive stuckness may need clarification, representation support or a prompt. Speed itself is not the target unless the task says it is.

Follow-up questions should expose, not upgrade, the answer

A follow-up is strongest when it reveals what is already in the learner’s model. It becomes misleading when the wording quietly supplies the missing concept. The Tutor Handbook’s Follow-Up Question owns the technical boundary; this public route shows what that means in teaching.

Learner response	Weak follow-up	Better follow-up
“The character is nervous.”	“Yes, because he keeps looking at the door, right?”	“Which detail made you infer nervousness?”
“Use division.”	“Because we are sharing equally?”	“What relationship in the problem makes division fit?”
“The plant grew less because it had less light.”	“So less light means less photosynthesis?”	“What changed, and how could that change affect the plant?”
“This answer is wrong.”	“Check the sign in line 3.”	“Which step could you verify independently?”

The better follow-up leaves the missing relationship with the learner. If the learner cannot produce it after a reasonable attempt, the tutor may need to teach directly rather than continue interrogating.

Know when to stop questioning and teach

Questioning is not inherently more sophisticated than explanation. Once the tutor has enough evidence that a relationship or prerequisite is missing, continued probing can waste time and increase frustration. The teaching system should switch mode.

A useful stopping rule is: if further questions are no longer likely to change the diagnosis, teach the missing relationship. After teaching, use a fresh question to check whether the learner can now use it.

Question return: answer, clarify, prompt or hand back

Learner questions also create a decision. “Which formula do I use?”, “Is this evidence okay?”, “What should I revise?” and “Can you check this?” can be genuine requests for missing knowledge or attempts to outsource a decision the learner needs to practise. The Question Return Gate owns the professional rule.

The public teaching version uses four responses: answer directly when knowledge is genuinely missing; clarify when the task is ambiguous; give bounded support when the learner has a partial route; or return the decision when the learner has enough information to practise choosing.

Learner asks…	Tutor decision
“What does this word mean?” and the word is prerequisite vocabulary not under assessment	Answer or clarify efficiently.
“Which formula?” after the learner has been taught the selection cues	Return part of the decision: “What quantities and relationship do you have?”
“Is my evidence enough?” in an argument lesson	Ask the learner to state the claim and explain what the evidence proves.
“Can you check this?” before any self-check	Return to the agreed verification routine first.
“I don’t know what the question means.”	Clarify task language before judging the academic capability.

Questioning in a three-student tutorial

Three learners create a special evidence problem: one answer can leak the route to the other two. If the fastest learner always speaks first, the tutor loses independent evidence from everyone else. Good small-group questioning therefore needs response architecture, not just interesting questions.

Think first, then share

Give all three learners a short independent think/mark/write period before inviting responses. This preserves each learner’s first model. The tutor can then compare rather than only hear the quickest answer.

Use sideways follow-ups

After one learner responds, do not immediately ask “Does everyone agree?” Ask another learner to compare: “What is the same in your reasoning? What is different?” or “Can you find a case where that rule would fail?” This keeps the second learner intellectually active rather than turning them into an audience.

Protect the original learner

If Learner A gives a partly correct answer, avoid letting Learner B repair it completely before A has had a chance to respond to one bounded follow-up. Peer explanation is valuable, but it should not erase evidence about what A can now carry.

Rotate question jobs, not just turns

Equal turns do not guarantee equal learning. One learner may need a retrieval question, another a representation question, another a transfer question. The target can be shared while the diagnostic job differs. Rotation should follow learner state and evidence rather than a mechanical speaking order.

Questioning by learner state

State	Best questioning job	Avoid
Blocked	Reveal first usable move; clarify target; test prerequisite.	Repeated “why?” questions when learner lacks the relationship.
Fragile	Retrieve after delay; reduce cues; ask learner to reconstruct route.	Near-identical questions immediately after teaching.
Stable	Change representation or context; ask learner to choose method.	More topic-labelled execution that supplies selection.
Transfer-Ready	Mix cases; ask for comparison, boundary and counterexample.	Treating one unusual question as proof of full transfer.
Examination-Ready	Ask paper-level planning/checking/recovery questions.	Reteaching stable content instead of testing integration.

Questioning should connect to the Capability Profile

A questioning episode often produces the observations that feed a capability profile. Record the task, support, wait condition, learner response, first weak link and what remains uncertain. Use A Capability Profile Without Labelling the Child when the evidence needs to be assembled across tasks without turning one answer into identity.

This also prevents questioning records from becoming a catalogue of defects. The profile should change when fresh evidence changes.

Accessible questioning without silent target changes

The route into a question may need to change. A learner may need more processing time, clearer visual presentation, a typed response, rephrased procedural instructions or another access condition. Use Accessible Learning Tasks Without Silent Target Changes to preserve the target while changing access.

Rephrasing must be handled carefully. If a Mathematics question asks the learner to infer a relationship from wording, rephrasing that wording into the relationship changes the construct. If the difficulty lies in an irrelevant procedural sentence, clarification may improve access without changing the mathematical target.

Student-generated questions reveal ownership

One of the strongest questioning outcomes is that learners begin generating useful questions themselves. This is not simply “ask more questions”. Good learner questions identify uncertainty, test an assumption, seek evidence, compare alternatives or plan a next move.

Weak learner question	Stronger learner question
“I don’t get it.”	“I understand the first step, but I don’t know how to decide whether these two quantities form a ratio or difference.”
“Is this right?”	“I substituted my answer back into the equation and both sides match; is there another condition I should check?”
“What do I write?”	“My evidence supports the claim, but I am not sure how much explanation the question asks for.”
“Which one is the independent variable?”	“I know what I changed and what I measured; I am unsure whether this third condition needs to stay constant.”

Teaching learners to ask these questions transfers diagnostic work toward the learner. The long-term aim is not a classroom full of tutor prompts; it is a learner who can locate uncertainty and choose a productive question before asking for help.

Questioning and Assessment Evidence

Answers are evidence under conditions. One correct response after heavy prompting is different from one generated independently after delay. One wrong response on a linguistically inaccessible task may not identify the academic weak link. Use How Assessment Evidence Works to interpret what the questioning episode can legitimately show.

A useful record might say: “On an unfamiliar ratio problem, learner identified the whole after one generic prompt and independently chose the operation; representation selection remains untested without prompt.” That is more informative than “understood after questioning”.

Fresh-question retests: questioning is not complete until the learner can act without the conversation

Questioning can create the illusion of progress because a skilled tutor can guide a learner to a strong answer in real time. The conversation may be educationally valuable, but it is not yet proof that the learner can reconstruct the reasoning alone. A questioning route therefore needs a fresh-question retest.

The fresh task should preserve the underlying capability while removing the specific conversational path that helped the learner. If the tutor used “What is the whole?” to unlock a fraction problem, the retest should contain a new context where the learner must identify the whole independently. If a follow-up helped an English learner narrow answer scope, the retest should use a new passage and question demand.

Questioning episode	Fresh acceptance check
Tutor asks learner to compare two fraction models and notice the changing whole.	New fraction problem with a different story and no comparison prompt.
Tutor probes why one piece of evidence supports an inference.	New passage with a different inference and no “why?” scaffold.
Tutor asks what variable was deliberately changed in an investigation.	New investigation design where learner labels changed/measured/controlled conditions independently.
Tutor asks which algebra step could be verified.	New equation where learner chooses and applies a verification method without being told to check.
Tutor uses wait time until learner identifies method.	Later mixed task where tutor remains neutral and no prompt arrives at the familiar latency.

Use Learning Practice & Review to choose the next fresh task and How We Know Learning Has Really Held to test delay, variation and transfer. The conversation is part of teaching; the fresh task is what lets the system update the learner state.

Questioning casebook: English

Case 1: literal answer, wrong question scope

A Primary 5 learner copies a true sentence from the passage but the question asks why the character acted that way. The weak follow-up is “What does that tell us about his reason?” because it nearly gives the inference. A better sequence asks: “What does the question want—an action, a reason or a feeling?” then “Which detail helps you answer that exact demand?” If the learner can now identify scope but still cannot infer, the tutor has separated task reading from inference.

Fresh check: new passage, learner underlines the demand word and writes one evidence-bounded answer without tutor dialogue.

Case 2: long essay, unclear argument

A Secondary 3 learner gives many examples but the argument remains weak. Instead of “Why is this example relevant?”, which may trigger a generic answer, ask “What claim is this paragraph trying to prove?” If the learner cannot state the claim, the problem is upstream of evidence explanation. If the claim is clear, ask “What does this example add that the previous one did not?” The questioning locates whether the weakness is claim control, evidence selection or explanation.

Fresh check: learner plans a new paragraph by writing claim → reason → evidence purpose before drafting.

Case 3: grammar editing

A learner identifies that a sentence “sounds wrong” but cannot classify the error. Ask first for what relationship is broken: subject–verb, tense sequence, reference, modifier position or sentence completeness. Then ask for one minimal correction. This keeps error detection separate from rewriting the whole sentence.

Questioning casebook: Mathematics

Case 4: word problem method selection

A Primary 4 learner immediately multiplies because the previous five questions used multiplication. The tutor should not ask “Are you sure it is multiplication?” because that signals error without revealing why. Ask “What two quantities are related, and what does the question want from them?” Then change the numbers but preserve the relationship. If method changes appropriately, the learner is beginning to select rather than pattern-match.

Case 5: algebra sign error

A Secondary 2 learner writes a correct first line and an invalid sign change in the second. Instead of pointing to the sign, ask “What operation happened to both sides between these two lines?” If the learner cannot name one, the shorthand is hiding equality. Teach the equality transformation, then retest on a new equation.

Case 6: geometry recognition

A learner knows angle facts but misses them in a rotated diagram. Ask “Which relationships remain true if I turn the page?” or “What information comes from the figure, and what comes from the geometry rule?” The question reveals whether the learner depends on orientation or can identify invariant structure.

Questioning casebook: Additional Mathematics

Case 7: identity selection

A Secondary 4 learner knows several trigonometric identities but does not know which to use. Avoid “Could you use the double-angle identity?” Ask “What form would make the two sides comparable?” or “Which expression is currently preventing you from solving for the angle?” This reveals whether the learner can work backward from a useful form.

Case 8: differentiation plus algebra

A learner differentiates correctly but loses marks during simplification. Ask “Where does calculus end in your working?” This separates derivative knowledge from algebraic manipulation. Then the next owner can be algebra repair rather than more differentiation practice.

Questioning casebook: Science

Case 9: keyword without mechanism

A Primary 6 learner writes “conduction” but no causal chain. Ask “What changed first?” then “How does that change move through the system?” The sequence forces relationship before vocabulary decoration. If the learner still cannot explain transfer, teach the mechanism rather than asking for more keywords.

Case 10: overclaim from data

A learner sees one bar higher than the others and says “this method is best”. Ask “Best for which measured outcome?” then “What does the graph not tell us?” This reveals whether the learner can separate observed pattern from broader judgement.

Case 11: fair test

A learner lists many “controlled variables” indiscriminately. Ask “Which of these could create another explanation for the result if it changed?” That question connects control to causal interpretation rather than memorising a list.

Case 12: prediction after seeing results

A learner writes a “prediction” that matches the revealed outcome. Ask “What would you have written before the data existed?” then require the next investigation to record prediction first. Questioning exposes the reasoning stage rather than only the vocabulary label.

Questioning in examination preparation

Examination questioning should not turn a full paper into a tutor-led conversation. Once paper practice begins, the learner needs enough uninterrupted work for timing and method selection to become visible. Use questions mainly during review, or at pre-planned checkpoints, rather than continuously rescuing the paper.

Useful review questions include: “Where did time first begin to drift?”, “Which question should you have left earlier?”, “Which error could your checking routine realistically have caught?”, “Which topic error was actually a representation error?”, and “What will you do differently on the next fresh paper?”

These questions route to Examination Craft when the problem is paper execution rather than subject knowledge.

Questioning in homework review

Homework review should not begin with the tutor correcting every wrong answer. Start with the learner’s own evidence: “Which question felt least certain?”, “Where did you change your mind?”, “Which answer did you check?”, “Where did someone help you?”, “Which problem could you now explain without the worksheet?”

These questions surface support conditions and metacognition. They also help distinguish completed homework from learning that has actually held.

Questioning and misconceptions

Misconceptions are not best repaired by telling the learner they are wrong repeatedly. Ask a prediction that the misconception and correct model would answer differently. Then compare the result with the learner’s expectation. The question becomes a test of the model.

For example, if a learner believes a larger denominator means a larger fraction, compare 1/3 and 1/8 using equal wholes and ask which piece is larger before calculating. If a learner believes heavier objects always fall faster in ordinary school reasoning, design a safe evidence discussion that distinguishes mass from the relevant conditions. The goal is not to win an argument but to make the model testable.

Questioning and uncertainty

Learners should be allowed to say “I am not sure” without the tutor treating uncertainty as failure. Good follow-ups can calibrate uncertainty: “What part are you sure about?”, “What two possibilities are you choosing between?”, “What evidence would decide?” This turns uncertainty into a reasoning state rather than a confidence defect.

A learner who can locate uncertainty precisely is often closer to independence than one who gives a confident unsupported answer.

Questioning and self-checking

Self-checking improves when the question becomes internal. Instead of the tutor repeatedly asking “Are you sure?”, teach the learner a small checking question tied to the task: “Does the size make sense?”, “Do both sides remain equal?”, “Did I answer the exact question?”, “Does my conclusion go beyond the data?”

The endpoint is not the tutor asking better checking questions forever. It is the learner using the relevant question before feedback arrives.

Questioning and planning

Planning questions should expose decisions without completing them. For a composition: “What must change by the end?” For Mathematics: “What information do you need before you can calculate?” For Science: “What would you measure to answer the question?” For revision: “Which weak link would affect the most later tasks?”

If the tutor supplies the plan, the learner can execute without learning to plan. The fresh retest should therefore require independent planning on a new task.

Questioning and retrieval

Retrieval questions should require recall before recognition where recall is the target. Showing four options can make a blocked memory look stronger. Conversely, recognition can be a useful intermediate step during repair. Record which question form was used so the capability profile does not overstate retrieval.

After successful retrieval, ask one connection question only if it serves the next job. Questioning should not turn every fact into a long discussion that dilutes practice density.

Questioning and transfer

Transfer questions change one meaningful condition. “Would your method still work if the whole changed?”, “What if the graph used a different scale?”, “What if the evidence contradicted your first claim?”, “Which part of your rule survives?” The changed condition should test the same underlying capability rather than introduce an unrelated trick.

Questioning and learner confidence

Do not use leading questions merely to manufacture successful answers and then call the learner confident. Accurate confidence comes from evidence that the learner can solve, explain and recover. Questions can help the learner notice that evidence: “What did you do independently this time that needed a prompt last week?”

Questioning and parent conversations

Parents can use one diagnostic question before giving advice. Instead of “Why did you get this wrong?” try “Where did the question first stop making sense?” Instead of “Did you study?” ask “What could you retrieve without notes?” Instead of “Do you understand now?” ask “Show me one new example you can do without the correction beside you.”

The parent should then allow the answer to change the response. A useful question whose answer is ignored becomes another ritual.

Questioning with AI-generated prompts

AI can draft question variants, contrast cases and practice prompts, but it cannot decide automatically which learner uncertainty matters. The tutor must specify the question job, review correctness, control difficulty, protect privacy and decide what evidence the answer will support.

• Ask AI for several candidate questions serving one explicit diagnostic job.
• Verify subject accuracy and age/syllabus fit.
• Remove wording that gives away the answer.
• Choose one question that distinguishes plausible learner models.
• Record if AI-generated scaffolding was used in the learner attempt.
• Use a fresh independently selected task before changing learner state.

The AI is a drafting tool. The educational judgement remains with the human tutor and the learner evidence.

A questioning audit after the lesson

Audit prompt	What to look for
Which question changed my teaching decision?	Keep questions that actually altered the next move.
Which question produced only performance theatre?	Remove questions that invited chorus answers or predictable agreement without evidence.
Did I wait long enough?	Check whether prompts arrived before learners had a real chance to think.
Did a follow-up give away the route?	Rewrite it around evidence, comparison or explanation.
Did one learner answer for the group?	Add think-first or individual response structure next time.
Did I keep questioning after diagnosis was already clear?	Teach sooner once the missing relationship is known.
What fresh evidence will accept the repair?	Reserve a new task; do not rely on the successful conversation.

Common questioning failure modes

• Guessing funnel: a chain of increasingly leading questions until the learner says the tutor’s answer.
• Empty why: repeated “why?” when the learner lacks the prerequisite knowledge to explain.
• Speed bias: fastest response is treated as strongest understanding.
• Chorus evidence: group agreement replaces individual evidence.
• Answer leakage: one learner’s full solution removes the diagnostic opportunity for peers.
• Revoicing upgrade: tutor paraphrases a weak answer into a strong one and then credits the learner with the stronger reasoning.
• Question overload: too many prompts fragment the learner’s own line of thought.
• Ritual questioning: tutor asks but never changes teaching based on the answer.
• Identity inference: one response becomes “weak”, “shy”, “careless” or “bright”.
• No retest: guided success is recorded as independent learning without fresh evidence.

Frequently asked questions

Are open questions always better than closed questions?

No. A closed question can efficiently distinguish two interpretations, while an open question can produce vague language that hides reasoning. Choose the form from the evidence job, not from a hierarchy of “higher-order” versus “lower-order” question shapes.

How long should I wait after asking?

There is no universal number. Wait long enough for the task and learner to generate evidence, and watch what happens during the pause. Retrieval of a simple fact and construction of an unfamiliar explanation need different latency. The technical boundary is handled in the Wait-Time and Prompt-Latency Gate.

Should I call on students randomly?

Response selection is a teaching-design decision. Preserve thinking time and psychological safety, and avoid using unpredictability merely to police attention. In a three-student tutorial, think-first responses and planned comparison often provide better evidence than surprise calling.

What if the learner says nothing?

First distinguish processing time, unclear task language, missing knowledge, access barrier and avoidance. Clarify or teach where needed. Silence alone does not identify the cause.

What if the learner gives the right answer immediately?

Ask whether the decision requires more evidence. Sometimes a right answer is enough and the lesson should move on. If reasoning or transfer matters, use one discriminating follow-up or changed case rather than interrogating every correct answer.

Should tutors always ask for explanations?

No. Explanations are useful when reasoning is the target or when the tutor needs to distinguish models. They can reduce practice density when overused. Automatic facts, fluent procedures and timed work sometimes need execution rather than continuous verbalisation.

Can peer questioning help?

Yes, if the peer question preserves the other learner’s thinking rather than becoming answer delivery. Teach question jobs such as clarification, evidence, comparison and boundary. Protect independent evidence before discussion where it matters.

How do I know a questioning strategy worked?

Not because the conversation sounded deep. Look for a better fresh attempt, more precise self-questioning, reduced prompt need, stronger transfer or a clearer teaching decision. Questioning is successful when it improves learning or evidence.

Can questioning become too supportive?

Yes. Leading questions can carry the reasoning. That is why prompts and follow-ups should be recorded as support conditions and later faded or removed where independence matters.

How does questioning connect to accessibility?

Access conditions can change how a learner enters the question without changing the target. More processing time, clearer layout or a different response mode may be valid. Rephrasing that supplies the missing relationship can change the construct. Use the accessibility owner for that boundary.

What is the endpoint of good questioning?

The learner increasingly asks useful questions internally and externally: What is being asked? What do I know? Which representation fits? What evidence supports this? Where could the error be? What should I try before asking for help? The tutor’s questions have transferred into learner control.

The WEP119 operating standard

Questioning that makes learner thinking visible follows one loop: decide the evidence needed → ask the smallest discriminating question → preserve thinking time → interpret the response cautiously → follow up without feeding the route → teach when diagnosis is sufficient → reserve fresh evidence → update the capability profile and next move.

The public owner is this page. Technical professional depth stays in the Tutor Handbook’s Wait-Time and Prompt-Latency Gate, Follow-Up Question, Question Return Gate, learner-talk and three-learner orchestration owners. Subject pages provide domain examples. The acceptance test is not a good discussion; it is better learner-owned performance on fresh work.

Questioning across school transitions

Primary 1: questions should reveal whether the learner can enter the task

In Primary 1, the first job is often access and interpretation rather than sophisticated verbal explanation. Ask short questions that reveal whether the child understands the instruction, recognises the representation and can make a first move. Long chains of “why?” can overload language before the underlying idea is visible.

Useful questions include: “What are you trying to find?”, “Which picture matches the sentence?”, “Show me where you will start,” or “Which two things belong together?” The tutor should observe whether the child can act, not only speak.

Primary 3: questions should separate new subject language from new ideas

Primary 3 introduces more formal Science and more complex task language across subjects. A learner may know the everyday idea but not the school term, or know the term without the relationship. Question both separately: “What did you observe?” before “What scientific word describes that?”, or “Show the relationship” before asking for a formal explanation.

Primary 5: questions should find the dependency before the workload expands

Primary 5 errors often sit at the intersection of older foundations and newer multi-step demands. Ask questions that locate the first dependency: “What is the whole?”, “Which information changes after the first step?”, “What does the question ask you to compare?”, “Which Science relationship connects these two observations?”

The aim is to avoid treating every lower mark as a broad Primary 5 weakness.

Primary 6: questions should begin transferring into examination control

By Primary 6, questioning should increasingly include paper decisions: “Which question would you leave and return to?”, “What checking method fits this answer?”, “Which topic cue is absent here?”, “What would you do if this first method does not work?” The tutor is preparing the learner to ask those questions internally during PSLE conditions.

Secondary 1: questions should expose transition load as well as subject knowledge

Secondary 1 performance can fail because the learner has lost the weekly operating system. Ask about task location, start routine, subject expectations and help-seeking alongside academic questions. “What do you need to know?” and “What do you need to organise?” may have different owners.

Secondary 3: questions should distinguish abstraction from overload

As abstraction rises, ask whether the learner understands the new representation, the underlying prerequisite and the weekly capacity needed to practise it. A learner who says “I hate A-Math” may need algebra repair, not a motivation lecture. A learner who knows the work but cannot sustain the week may need runtime repair.

Secondary 4: questions should compress toward high-leverage decisions

The final year is not the time for endless exploratory questioning. Ask questions that change the next practice block: Which weak link still controls several marks? Which error is knowledge and which is paper execution? What fresh evidence shows that the repair held? What can now be left alone?

The questioning handover

A useful questioning sequence ends by handing something back to the learner. The learner should leave knowing the question they can ask themselves next time. That question becomes a portable control.

• Mathematics: “What relationship is this really asking about?”
• English comprehension: “What exactly is the question demanding?”
• Science: “What changed, what mechanism connects it, and what was observed?”
• Writing: “What is this paragraph trying to prove or change?”
• Checking: “What verification method matches this kind of answer?”
• Help-seeking: “Where exactly does my route stop?”
• Revision: “Which weak link is controlling the most later work?”

When the learner begins using these questions without the tutor, questioning has crossed from teaching technique into independent learning.

The final evidence review

Before updating a learner state because of a questioning episode, ask four things: Was the task fresh enough? How much prompting was used? Did the learner reconstruct the reasoning or merely agree with the tutor’s wording? Did the capability survive a later task without the same conversational support?

If those answers are unclear, the correct state may be “not yet enough evidence”. This is preferable to overstating mastery or weakness.

Questioning as part of the wider eduKateSengkang runtime

Questioning sits between observation and action. Start Here supplies the symptom. Assessment Evidence clarifies what the response can support. The Capability Profile holds the learner-specific model. Accessible Learning protects the target when conditions change. Subject owners repair the knowledge. Learning Practice & Review supplies the fresh task. Learning Has Held checks delay and transfer. Examination Craft takes over when the final problem is paper execution.

That architecture prevents one teaching technique from becoming a philosophy for every problem. Sometimes the right move is a question. Sometimes it is explanation, modelling, practice, silence, feedback, recovery or simply time to let the learner work.

The minimum questioning implementation checklist

Before a questioning routine becomes part of normal tutoring, check whether it protects learner thinking rather than merely increasing tutor talk. The checklist is deliberately short enough to use after a real lesson.

• Purpose: Can the tutor state what evidence each important question was meant to produce?
• Discrimination: Did the question distinguish plausible learner models, or could any answer have been interpreted the same way?
• Wait: Did the learner have enough time to think before the tutor rephrased, hinted or answered?
• Access: Were task-language or response barriers separated from the academic target?
• Follow-up: Did the follow-up reveal reasoning without supplying the missing route?
• Group evidence: In a three-student lesson, did each learner have an independent thinking opportunity before answer leakage?
• Stop rule: Did the tutor stop questioning once the missing relationship was clear and teach it directly?
• Fresh acceptance: Was there a later task where the learner acted without the diagnostic conversation?
• Record: Were prompts/support conditions visible enough to interpret the result honestly?
• Handover: Did the learner leave with a useful self-question they can apply next time?

If the routine passes those checks, questioning is functioning as part of the learning runtime: it reveals what matters, protects evidence, changes the next move and eventually transfers more diagnostic control to the learner.

Questioning versus explaining: the final boundary

A questioning system fails when the tutor assumes every gap should be discovered through more questions. Some knowledge is genuinely missing and should be taught clearly. The diagnostic value of questioning is to show when explanation is needed and what explanation should contain.

If the learner lacks prerequisite vocabulary, has never seen the representation, or cannot generate the relationship after sufficient opportunity and one or two discriminating prompts, direct teaching is often the more efficient and humane move. Explanation should then be followed by a fresh learner attempt so that the tutor can see whether the new knowledge has become usable.

The opposite error is explaining too soon. If the learner already has the knowledge but needs time to retrieve, select or articulate it, immediate explanation replaces evidence with tutor performance. The practical boundary is simple: question while the answer can still reveal a meaningful learner decision; explain once the missing knowledge is sufficiently clear; then retest without the explanation doing the work.

That boundary keeps questioning inside a complete teaching cycle rather than turning it into a style preference. The aim is not to ask the most questions. It is to ask the questions that make the next educational move more accurate and eventually less necessary.

One last operating rule

Every important question should either reveal evidence, improve the learner’s next move, or transfer more control to the learner. If it does none of those jobs, remove it and give the learner time to work.

Atlas return: this page keeps its own teaching or learning job. Return to Learning Atlas V2.0 when the next decision is learner state, subject route, parent action, examination performance or the next evidence check.