How Checks for Understanding Work | Why Evidence Must Change the Next Teaching Move

A teacher asks, “Does everyone understand?” Thirty faces look back. A few students nod. Nobody volunteers a problem. The lesson moves on.

Five minutes later, the independent task reveals that half the class has misunderstood the same idea.

The failure did not happen because the teacher failed to care. It happened because “Does everyone understand?” produced almost no diagnostic evidence. Students had to judge their own understanding, decide whether it was safe to admit uncertainty, and then interrupt the lesson publicly. Silence looked like understanding because the question was designed in a way that made misunderstanding easy to hide.

Checks for understanding work differently. They are deliberate attempts to elicit evidence of what students can currently retrieve, explain, discriminate or apply so the teacher can decide what to do next. The important word is not check. It is decide. If the evidence never changes instruction, the activity has become a ritual rather than formative teaching.

The Australian Education Research Organisation’s current Monitor Progress practice guide, updated in May 2026, describes checking for understanding as a way to determine what students know and can do, identify gaps, and adjust teaching through additional instruction, guidance or feedback. AERO’s 2025–2026 classroom videos emphasise frequent checks, varied participation methods, prompt response to errors, guidance after unsuccessful attempts, and use of student responses to evaluate the teaching itself. The Education Endowment Foundation’s recent work on checking for understanding makes a closely aligned point: effective checking is not a collection of techniques; it begins with evidence anchored to learning intentions, gathered from more than volunteers, and then used to guide what happens next. EEF’s wider feedback evidence also stresses that feedback depends on first assessing what needs to improve.

This article owns one narrow canonical mechanism on eduKateSG: the evidence-to-decision loop of checking for understanding during instruction. It does not own questioning generally, all-pupil response systems, classroom scanning, teacher noticing, mini-whiteboards, feedback, or formative assessment as a whole. Those are neighbouring tools and systems. A check for understanding has one specific job: produce evidence that is good enough to justify the next teaching move.

The 50-second answer

A useful check for understanding has six stages:

Name the learning claim. What exactly should students know or be able to do at this point?
Design a prompt that can reveal that learning. Ask students to retrieve, discriminate, explain, represent or apply—not merely report confidence.
Collect evidence from enough of the class. One confident volunteer is not a sample of thirty learners.
Interpret the pattern. Is the error isolated, widespread, procedural, conceptual, linguistic, or caused by the task itself?
Act on the evidence. Continue, reteach, model, give a smaller prompt, change the example, regroup, or provide targeted support.
Check again. A correction is not complete until the teacher knows whether the new explanation changed understanding.

The simplest useful formula is:

claim → evidence → interpretation → response → recheck.

The technique can be a question, a mini-whiteboard, a short written response, an example/non-example choice, a worked step, a diagram, a gesture, a digital poll, or a student explanation. The technique is not the mechanism. The mechanism is using visible evidence to control the next instructional decision.

1. The check must be tied to a specific learning claim

A teacher cannot interpret a response unless they know what the response is evidence of.

“Any questions?” is broad. “Which of these two fractions is larger, and what comparison did you use?” is tied to a learning claim. “Show me the clause that gives the reason” is tied to a learning claim. “Predict what happens to current if resistance increases while voltage stays fixed” is tied to a learning claim.

Specificity makes the evidence interpretable.

Concrete example: the class is learning that multiplying by a number smaller than one can reduce a quantity. The teacher does not ask, “Are we okay with decimals?” Instead, every student chooses which is larger: 0.6 × 80 or 80, then gives one sentence of reasoning. The teacher can now see whether students still carry the misconception that multiplication always makes a number larger.

Caveat: a check can be specific but trivial. Asking students to repeat a definition may not reveal whether they can use the concept. The prompt should match the intended depth of learning.

2. Confidence is not the same as understanding

Students are poor instruments for some kinds of self-diagnosis, especially while knowledge is new. A learner may feel confused yet answer correctly with support. Another may feel certain because the explanation sounds familiar.

“Thumbs up if you understand” measures a mixture of confidence, social comfort, self-awareness and actual knowledge.

That does not make confidence data useless. It makes it a different variable.

Concrete example: after teaching direct and inverse proportion, the teacher asks students to rate confidence and then solve one diagnostic item. Several high-confidence students choose the wrong model. That discrepancy is useful: it reveals calibration problems, not merely content problems.

A check for understanding should usually ask the learner to do something with the knowledge.

3. Volunteers distort the picture of the room

When teachers ask a question and call on the first raised hand, the answer may be excellent and still tell them little about the rest of the class.

Volunteer responses are shaped by confidence, speed, personality, language, status and prior attainment. The students most willing to answer can become a biased sample.

All-pupil response methods are valuable partly because they widen the evidence base. But the deeper principle is representativeness, not a specific tool.

Concrete example: a teacher asks a multiple-choice conceptual question. Instead of taking one answer, every student holds up an answer card. The teacher sees that 40 percent chose the same distractor. The distractor now becomes evidence of a shared misconception.

Boundary: “all pupils respond” does not mean every response has equal diagnostic value. A poorly designed question can collect thirty pieces of weak evidence.

4. Good checks discriminate between plausible understandings

The best diagnostic prompt often includes an option that a partially informed learner would choose.

If every wrong answer is obviously absurd, the item tells the teacher only whether students can avoid nonsense.

A stronger check separates competing mental models.

Concrete example: students are learning why seasons occur. The teacher offers four diagrams: distance from the Sun, axial tilt with fixed orientation, daily rotation, and changing cloud cover. Each distractor maps to a plausible misconception. The pattern of choices tells the teacher what needs repair.

This is one reason examples and non-examples, hinge questions and carefully chosen distractors can be powerful.

Caveat: teachers should avoid over-interpreting a single wrong choice. A student may misread, guess or press the wrong button. Look for patterns and, when necessary, ask for reasoning.

5. The check should happen while there is still time to change the lesson

An assessment at the end of the unit can reveal misunderstanding, but the opportunity cost is high. Students may have practised errors for days.

Checks for understanding are most valuable when inserted before the next layer depends on the current one.

Concrete example: before moving from expanding single brackets to solving equations containing brackets, the teacher checks whether students can correctly expand three carefully selected examples. If the class is unstable, the lesson does not proceed as planned.

This may feel slower in the moment. It is often faster than repairing a larger structure built on a weak foundation.

6. Hinge points deserve stronger checks

Some moments in a lesson are structurally important. If students misunderstand them, everything that follows becomes unreliable.

These are hinge points.

The teacher should spend more diagnostic effort there than on minor details.

Concrete example: in a Science lesson, students must distinguish mass from weight before interpreting gravitational field strength. That distinction is a hinge. A two-minute diagnostic check may be worth more than ten quick recall questions on vocabulary.

Boundary: not every fact can be treated as a hinge. The teacher needs curriculum judgement to identify which ideas carry future learning.

7. The prompt should reveal thinking, not only the final answer

A correct answer can be produced by the wrong method. A wrong answer can contain useful partial reasoning.

Whenever method matters, the check should expose at least one step of thinking.

Concrete example: students solve 3(x + 4) = 21. Several reach x = 3. The teacher asks them to show the first transformation only. Some divide 21 by 3 before subtracting 4; others subtract 4 from 21. The final answer alone would not reveal the difference because one route may later accidentally converge.

In writing, the equivalent might be asking students to identify the evidence sentence they would use before writing the paragraph. In reading, it may be underlining the phrase that supports an inference.

8. Checks can be deliberately tiny

A check does not need to become a quiz.

One well-chosen item can tell the teacher whether a concept is ready to support the next step.

Concrete example: before moving on from subject–verb agreement, the teacher gives one sentence where the grammatical subject is separated from the verb by a plural noun. If students choose the verb based on proximity, the misconception is visible immediately.

Tiny checks are especially useful during explanations because they preserve flow.

Caveat: tiny checks sample narrowly. They should be repeated across a sequence rather than treated as proof of full mastery.

9. Mini-whiteboards work only when the question and scan are good

Mini-whiteboards are popular because they make many responses visible at once. But the board itself does not create formative assessment.

A weak prompt such as “write one thing you learned” may generate thirty pleasant but non-diagnostic statements. A strong prompt targets a known misconception or decision point.

The teacher also needs a scanning routine. If boards are raised for two seconds and the teacher notices only the front row, the system creates the appearance of evidence without actual interpretation.

Concrete example: students write the next line of an algebraic manipulation. The teacher scans specifically for sign errors and incorrect distribution, then chooses whether to continue or reteach.

10. Multiple-choice checks can be deeply diagnostic

Multiple choice is sometimes dismissed as shallow. Poor multiple choice is shallow. Well-designed options can encode competing reasoning patterns.

A good diagnostic item asks the teacher to predict why a student might choose each distractor.

Concrete example: after teaching area and perimeter, the options for a rectangle problem include the correct answer, the perimeter, an answer produced by adding instead of multiplying, and an answer that ignores units. The distribution of choices points to different problems.

Boundary: choice selection does not always reveal why the learner chose it. Follow-up explanation may be needed.

11. Short writing can expose understanding that oral questioning misses

Writing gives all students processing time and leaves a visible product.

A stop-and-jot can be a check for understanding when the prompt is diagnostic rather than reflective only.

Concrete example: “In one sentence, explain why increasing sample size can reduce random sampling error but cannot automatically remove bias.” The teacher samples responses and notices many students confuse bias with variability.

The class then receives a contrastive explanation before proceeding.

12. A check can ask for a representation rather than words

Some concepts are better revealed through diagrams, number lines, graphs, tables or models.

Concrete example: instead of asking students to define acceleration, the teacher gives a velocity–time graph and asks students to sketch the section representing constant positive acceleration. The representation reveals whether they connect the concept to slope.

Caveat: representation skill can itself be a barrier. A learner may understand the concept but not the requested notation. Interpret cautiously.

13. Errors should be safe enough to surface

If wrong answers are socially costly, students will hide them.

A classroom that values checks for understanding needs a culture where errors are treated as information rather than public identity.

This does not mean praising every error as wonderful. It means responding in a way that keeps evidence available.

Concrete example: when 12 students choose the same wrong option, the teacher says, “Good, this tells me the distinction is not stable yet. Let’s compare the two models.” The response focuses on the knowledge problem, not the students.

Boundary: safety does not require pretending errors have no consequences in high-stakes contexts. During learning, however, hidden errors are more dangerous than visible ones.

14. The teacher needs a decision rule before the check

Teachers can collect evidence and still struggle to act because they did not decide what different patterns would mean.

A simple pre-lesson rule helps:

If nearly everyone succeeds, continue and increase challenge.
If a small minority struggle, provide targeted support while others continue.
If many students make the same error, pause and reteach.
If responses are scattered, inspect whether the question itself was unclear.

The thresholds do not need to be numerical laws. The point is to prepare possible responses before the live pressure of the lesson.

15. Reteaching should change something

Repeating the same explanation louder is not adaptation.

If the first explanation did not produce understanding, the second should change representation, example, level of guidance, language, sequence or amount of information.

Concrete example: students misinterpret equivalent fractions after a verbal explanation. The teacher switches to a number line and asks students to place 1/2, 2/4 and 4/8 at the same position. The new representation attacks the misconception differently.

Then the teacher checks again.

16. The recheck closes the loop

A common failure is to reteach and then move on because the second explanation felt clearer.

Clarity to the teacher is not evidence of student learning.

After repair, run a fresh item that tests the same underlying idea with different surface details.

Concrete example: after correcting a ratio misconception using red and blue counters, the teacher checks a new problem about map scale. If students succeed, the repair has stronger evidence of transfer.

17. Checks for understanding can reveal a teaching problem, not a student problem

When many students fail in the same way, the evidence may point back to instruction.

Perhaps the explanation assumed missing prior knowledge. Perhaps an example was ambiguous. Perhaps the task language created an irrelevant barrier. Perhaps the sequence moved too quickly.

AERO explicitly frames student responses as information teachers can use to check their own practice.

Concrete example: nearly every student chooses an incorrect interpretation of a graph. Reviewing the slide, the teacher realises the axis label was introduced after the data pattern. The error is partly instructional design.

This is professional evidence, not personal blame.

18. Checks should sample quiet learners deliberately

Some students look attentive, complete routines and rarely request help. They can disappear inside a lesson.

Whole-class response methods, short writing and classroom scanning protect against this invisibility.

Concrete example: one quiet learner consistently gives incorrect mini-whiteboard responses despite never asking a question. The teacher schedules a brief check during independent work and finds a foundational vocabulary gap.

Without systematic checking, compliance would have hidden the need.

19. Checks should not turn every minute into assessment

A classroom can become exhausting if students feel continuously tested.

The goal is not maximum evidence collection. It is enough evidence at high-value moments.

Teachers should protect stretches of explanation, discussion, reading, practice and exploration where constant interruption would be counterproductive.

Concrete example: during sustained silent reading, the teacher does not stop the class every three minutes for a comprehension poll. Instead, a targeted check occurs after a meaningful chunk.

Boundary: efficiency matters. Every check has an opportunity cost.

20. Digital polls are fast but can create false precision

Technology can collect thirty responses instantly and display percentages. That speed is useful. It can also make weak evidence look scientific.

A result of 73 percent correct says little unless the item was diagnostic and the teacher knows what the incorrect options mean.

Concrete example: a digital poll shows 70 percent choose option B. The teacher inspects the 30 percent and sees most chose D, a distractor linked to a known misconception. The pattern matters more than the percentage alone.

Caveat: device access, login friction and technical delay can make a digital check more costly than a paper or hand signal.

21. Checks should distinguish retrieval failure from concept failure

A student may understand an idea but fail to retrieve a term. Another may recall the term while misunderstanding the idea.

Diagnostic prompts can separate the two.

Concrete example: a student cannot recall the word “evaporation” but accurately describes liquid particles escaping into gas. The teacher supplies the label rather than reteaching the concept from zero.

Conversely, a student may correctly say “evaporation” while believing the water has disappeared. The label alone would hide the misconception.

22. Language can distort evidence in content subjects

A Science or Mathematics check can accidentally become an English test if the prompt uses unnecessarily complex language.

This matters especially for multilingual learners and students with language needs.

Concrete example: instead of asking, “Which statement most accurately characterises the proportional relationship represented by the graph?”, the teacher might ask, “Which sentence describes how y changes when x doubles?” if the linguistic complexity is not itself the learning target.

Boundary: academic language still needs teaching. Simplifying a diagnostic prompt should not mean permanently avoiding subject vocabulary.

23. Checks can reveal whether examples are transferring

Students often succeed when the practice item looks like the modelled example and fail when the surface changes.

A good check can deliberately vary context while preserving the underlying structure.

Concrete example: after teaching percentage decrease with sale prices, the teacher checks using population decline. If performance collapses, students may have learned the surface routine rather than the transferable relationship.

This is why checks for understanding can be designed as transfer probes, not only copies of the worked example.

24. Parents can use checks without becoming examiners

At home, the principle is similar: ask the learner to produce evidence rather than report confidence.

Instead of “Do you know this chapter?”, ask, “Explain the three stages without looking,” “Show me one example,” or “What mistake would someone make here?”

Keep the interaction low stakes. The goal is to reveal what needs review, not to turn dinner into a formal test.

Concrete example: a parent asks a child to explain why 3/5 is larger than 4/10 using any representation. The answer gives more information than “I did fractions already.”

25. Learners can learn to check themselves

Eventually, students should internalise part of the formative loop.

They can ask: What was I supposed to learn? What evidence do I have that I can do it? Where did I hesitate? What fresh example can I solve without notes?

This connects checking for understanding with metacognition, but the jobs remain distinct. The check generates evidence about current learning; metacognitive regulation uses that evidence to plan what to do next.

Concrete example: after revision, a learner chooses one unfamiliar problem rather than rereading notes. Failure on the problem triggers targeted review.

26. A good check can prevent unnecessary reteaching

Teachers sometimes reteach because the room feels uncertain. Evidence can show that the class is actually ready to move on.

Concrete example: a complex explanation feels shaky to the teacher, but a carefully designed application item shows nearly all students can transfer the idea. The teacher continues instead of spending ten more minutes repeating content.

Checks for understanding therefore protect time in both directions: they stop premature progression and unnecessary repetition.

27. The most important question is often “what will I do with each possible answer?”

Before asking a diagnostic question, the teacher can inspect the options and decide what each would imply.

If option A means the learner understands the invariant, B means a sign misconception, C means an order-of-operations error and D means random guessing, the teacher has a response map.

If the teacher cannot say what any wrong answer means, the question may be less useful than it looks.

This planning habit turns checks into instructional control rather than entertainment.

Worked case 1 — The nodding class

A teacher completed an explanation of simultaneous equations and asked whether everyone was comfortable. Several students nodded. Nobody objected. Independent work then stalled.

The next lesson used one diagnostic item before practice. Every student wrote the first elimination step on a mini-whiteboard. Nearly half added equations that did not eliminate a variable.

The teacher stopped and retaught the decision rule using colour coding and two non-examples. A fresh item showed much stronger performance.

The difference was not that the class suddenly became honest. The second lesson collected better evidence.

Worked case 2 — The perfect mini-whiteboard routine that changed nothing

A department used mini-whiteboards in every lesson and assumed formative assessment was embedded. Observation showed that teachers often asked recall questions, praised correct boards and moved on without interpreting error patterns.

The department changed one planning question: “What will each likely wrong answer tell us, and what will we do if we see it?”

Board use became less frequent but more diagnostic. Teachers began designing questions around hinge points rather than filling lesson plans with response routines.

The tool stayed the same. The thinking changed.

Worked case 3 — The student who knew the idea but not the word

A Primary Science pupil repeatedly left a vocabulary item blank. The teacher initially assumed the concept had not been learned.

A follow-up check asked the pupil to draw and explain what happened to water in an uncovered dish. The explanation was accurate. The missing piece was retrieval of the term “evaporation.”

The support shifted from full reteaching to vocabulary retrieval practice.

A more diagnostic check prevented unnecessary remediation.

Worked case 4 — The misleading correct answer

A Secondary student gave the correct answer to a percentage change question. When the teacher asked for the first step, the student revealed a method that would fail on a slightly different item.

The class then compared two methods and completed a transfer check using a different context.

The episode showed why final-answer accuracy can overstate understanding.

Practical route for teachers

Begin with the learning claim, not the tool. Decide what students should understand at the next hinge point and what response would provide credible evidence.

Predict likely errors. Choose a response method that gives you enough of the class to interpret the pattern. Build a decision rule: continue, target, reteach, change representation, or inspect the question itself.

During the lesson, collect the evidence and resist the urge to react to the fastest student. Scan the distribution. If needed, sample reasoning. Respond proportionately.

Then check again with a fresh item. Do not close the loop merely because your second explanation felt better.

Across a week, vary the evidence channel: oral explanation, short writing, diagrams, mini-whiteboards, multiple choice, worked steps, quick application. Different channels expose different aspects of understanding.

Practical route for learners

Do not ask only, “Does this look familiar?”

Ask for evidence from yourself.

Close the notes. Explain the idea. Solve one fresh problem. Draw the process. Identify an example and a non-example. State the common mistake. If you cannot do those things, you have found useful information.

A wrong answer during self-checking is not proof that study failed. It is a signal telling you where the next study minute should go.

After reviewing, check again with a new item. The recheck matters because recognition after seeing the answer can feel like learning when it is only renewed familiarity.

Practical route for parents and families

Replace “Have you studied?” with one or two low-pressure evidence questions.

Ask the child to explain a concept without notes, solve one unfamiliar example, or teach you the difference between two easily confused ideas. If the answer breaks down, help them identify the exact gap and return to the school materials.

Avoid rapid-fire testing unless the child finds it helpful. The purpose is diagnosis, not parental scoring.

For younger learners, drawing, sorting or demonstrating can reveal more than verbal interrogation.

28. Checking for understanding can improve pacing by making “slow down” and “speed up” evidence-based

Pacing is often treated as a personality trait of the teacher: fast teachers move quickly, careful teachers move slowly. A stronger model treats pace as a response to evidence. The class should move quickly through stable knowledge and spend more time where understanding is fragile.

Concrete example: a teacher planned fifteen minutes on revising linear graphs. A two-item check shows almost universal accuracy, including one transfer item. The teacher shortens the review and reallocates time to simultaneous equations, where evidence is weaker. On another day, the same check exposes a slope misconception and the review expands.

This prevents the timetable from becoming the hidden curriculum. The clock matters, but student evidence helps decide where time is educationally valuable.

Caveat: pacing cannot be infinitely adaptive. Teachers work within curriculum, timetable and examination constraints. Checks inform judgement; they do not remove those constraints.

29. Checks can separate a memory problem from a language problem from a reasoning problem

A wrong response is not a diagnosis. The next check can narrow the cause.

If a student cannot answer a Science question, the teacher might first simplify the language while preserving the concept. If the student then succeeds, language was part of the barrier. If not, the teacher might provide the key term and ask for the mechanism. If performance improves, retrieval was the bottleneck. If reasoning still fails, the concept itself may need reteaching.

This layered checking prevents teachers from treating every failure as the same kind of gap.

Concrete example: a learner cannot answer, “Explain why the rate of evaporation increases with temperature.” The teacher asks the student to point to the faster-moving particles in two diagrams. The student succeeds. A follow-up oral explanation reveals the learner understands particle motion but lacks the vocabulary “kinetic energy.” Support can now be precise.

30. Department-level checking can reveal curriculum weaknesses that no single lesson exposes

When teachers compare patterns across classes, checks for understanding can become curriculum evidence. If several classes make the same error at the same point, the problem may lie in sequencing, examples or prior-year assumptions rather than individual delivery.

A department might collect one common hinge item after a unit, not to rank teachers, but to see whether a concept is consistently unstable. The team can then inspect the curriculum materials, prerequisite knowledge and examples.

Boundary: common checks become harmful when used as covert performance scores. Their value lies in collective diagnosis and improvement, not simplistic comparison of teachers or classes.

Common failure modes

Asking “Does everyone understand?” and interpreting silence as evidence.
Taking one volunteer answer as representative of the class.
Using response technology without designing diagnostic questions.
Checking facts when the learning goal is application or reasoning.
Collecting evidence but continuing with the planned lesson regardless.
Reteaching with the same explanation that failed the first time.
Forgetting to recheck after correction.
Treating confidence as a direct measure of understanding.
Making errors socially costly so students hide them.
Checking so frequently that learning is constantly interrupted.
Using unnecessarily complex language that contaminates the evidence.
Interpreting one wrong answer as a stable misconception.
Looking only at final answers and missing faulty methods.
Using mini-whiteboards as theatre rather than information.
Assuming a high percentage correct proves durable learning.

Frequently asked questions

Is checking for understanding the same as formative assessment?

It is one important part. Formative assessment is broader. Checking for understanding focuses on eliciting and interpreting evidence during learning so instruction can adapt.

Is questioning the main way to check understanding?

No. Questions are one tool. Writing, diagrams, worked steps, sorting, mini-whiteboards, polls, models and demonstrations can all provide evidence.

How many students need to respond?

Enough to make the decision defensible. At important hinge points, whole-class or broadly representative evidence is usually stronger than one volunteer.

Should every wrong answer trigger reteaching?

No. Look for pattern, importance and cause. One isolated error may need a private prompt; a widespread misconception may need whole-class repair.

Are mini-whiteboards evidence-based by themselves?

No tool is effective merely because it is present. The value comes from the quality of the prompt, visibility of responses, interpretation and action.

What is a hinge question?

A hinge question is placed at a point where the teacher needs to know whether students have enough understanding to proceed to the next dependent step.

Can checks for understanding be graded?

They are usually most useful when low stakes, because students need to expose uncertainty. Grading can change behaviour and make the evidence less honest.

How do I avoid embarrassing students?

Use whole-class response methods, short private writing, neutral language about errors, and feedback that focuses on the idea rather than the person.

Sources and further reading

Australian Education Research Organisation, Monitor progress: https://www.edresearch.edu.au/guides-resources/practice-guides/monitor-progress
Australian Education Research Organisation, Monitor progress: Primary and secondary: https://www.edresearch.edu.au/guides-resources/videos/monitor-progress-primary-and-secondary
Education Endowment Foundation, How checking for understanding can guide your teaching: https://educationendowmentfoundation.org.uk/news/understanding-feedback-guides-teaching
Education Endowment Foundation, Feedback: The engine room of adaptive teaching: https://educationendowmentfoundation.org.uk/news/engine-room-of-adaptive-teaching
Education Endowment Foundation, Feedback: https://educationendowmentfoundation.org.uk/education-evidence/teaching-learning-toolkit/feedback
Education Endowment Foundation, Embedding Formative Assessment: https://educationendowmentfoundation.org.uk/projects-and-evaluation/promising-programmes/embedding-formative-assessment

Continue exploring on eduKateSG

The final idea

A check for understanding is not a pause in teaching. When it is designed well, it is teaching under feedback control.

The teacher makes a claim about what students should now know, asks for evidence that can test the claim, reads the pattern, changes the next move, and checks whether the change worked.

The most useful question is therefore not “Which checking technique should I use?” It is: What evidence would make me teach the next five minutes differently?