“My child understands in class but fails the test” is one of the most important symptoms in learning. It tells us that performance inside instruction and performance without instruction are not the same thing. The student may genuinely understand while the teacher is explaining — and still be unable to retrieve, select, transfer or execute the learning later.
Parents often describe this as a mystery: homework looked fine, tuition looked fine, the child could answer when asked at home, then the test score collapsed. The wrong conclusion is that the child “didn’t really understand anything”. The better question is what support was present during learning that disappeared during the test?
At eduKate Sengkang, this symptom is routed through the learner-state system. A student can appear strong in class while still being Fragile, reliable on familiar exercises but only Stable, capable of transfer but not yet examination-ready. The test is useful because it exposes which condition the learning cannot yet survive.
Five common reasons understanding does not become marks
| Cause | Classroom success | Test failure |
|---|---|---|
| Prompt dependence | Teacher questions lead the route. | No prompt arrives. |
| Retrieval fragility | Notes/examples remain nearby. | Knowledge must be produced cold. |
| Method-selection gap | Topic is already known. | Mixed paper hides which method applies. |
| Transfer gap | Practice resembles examples. | Wording or representation changes. |
| Execution gap | Time and support are forgiving. | Clock, sequencing and checking matter. |
The first diagnostic question
Ask the child to redo one test item several days later without the correction visible. Then vary the conditions. If they can solve it immediately after being reminded of the topic but not cold, retrieval is fragile. If they solve the original but fail a changed version, transfer is weak. If they solve both untimed but fail under a paper clock, examination execution is the likely next target.
Why classroom understanding can be completely genuine
A teacher or tutor provides structure that is easy to underestimate: sequence, emphasis, examples, vocabulary, worked steps, immediate feedback and cues about what matters. The learner may follow that structure well. The test asks for a different capability: reconstruct the structure from the question itself.
The educational mistake is not that explanation was useful. Good explanation is essential. The mistake is ending the learning cycle at “that makes sense”. Understanding must later be retrieved and used without the explanatory scaffold.
The four-test ladder
- Explain-back: Can the learner explain the idea in their own words immediately?
- Closed-source retrieval: Can they reconstruct it with notes closed?
- Fresh-item transfer: Can they use it on a new example or representation?
- Paper condition: Can they still do it mixed with other material and under time?
Where the ladder breaks tells you the next job.
English
In English, classroom understanding can be supported by discussion. A learner understands a passage after the tutor asks the right questions, sees why a composition paragraph works after it is annotated, or recognises a grammar rule when the exercise title announces it. Tests remove those supports.
Retest with a new passage, no leading questions and a demand for evidence. In writing, ask the learner to plan a fresh response before seeing a model. For language depth, use the English Learning Hub or SETC.
Mathematics
In Mathematics, the class often works chapter by chapter. This reduces method-selection load because the learner knows what technique is likely to be relevant. A test mixes chapters. Now the student has to diagnose the problem before solving it.
Use mixed sets before the test, but only after the underlying methods are stable. Ask for the method choice and reason before calculation. If the student cannot start even familiar work, return to the Blocked route. If the method disappears after delay, use Fragile.
Science
Science discussions can create a strong feeling of understanding because explanations are co-constructed. A test requires the learner to identify the mechanism from a new setup and write the causal chain independently. Memorised keywords may not be enough.
Ask for prediction and explanation from a fresh diagram. If the scientific concept itself is missing, route to the Science Hub or the canonical knowledge owner at eduKateSingapore.
A-Math
A-Math can create false confidence when examples are strongly patterned. The student follows differentiation, trigonometry or functions while the chapter title tells them what to do. A mixed test removes the title and exposes dependency gaps.
Keep small mixed retrieval and method-selection checks throughout the term rather than waiting until the full paper.
How parents can test without reteaching
- Ask for one fresh question, not the exact test item first.
- Keep notes closed for the first attempt.
- Do not tell the child which chapter it belongs to.
- Ask “What tells you this method fits?”
- If stuck, give the smallest cue that could restart the route.
- Record whether the cue was about knowledge, method, representation or confidence.
- Retest later with less support.
When homework success is misleading
Homework often arrives immediately after teaching, in a predictable sequence and with resources nearby. That makes it excellent for practice but a weak standalone measure of independent learning. A child can complete homework correctly because the environment contains the answer route.
Add one “cold” item from older learning and one mixed item without a chapter label. The goal is not to make homework harder; it is to collect better evidence.
The test review that separates causes
| If the child… | Likely route |
|---|---|
| Cannot recall the core idea | Fragile learning / retrieval |
| Knows methods separately but chooses badly | Stable → transfer / selection |
| Solves changed questions poorly | Transfer gap |
| Solves untimed but not timed | Examination-ready work |
| Cannot begin familiar work | Blocked / prerequisite diagnosis |
| Makes local slips despite correct plan | Execution / checking routine |
What tutors should change
Do not respond to every low test by adding more explanation. If explanation already produces strong in-class performance, the missing stage may be retrieval, variability or paper execution. Teaching should move to the point where independence actually breaks.
This is also why How Practice Works matters: practice must preserve the learner’s responsibility to retrieve and choose.
A two-week repair cycle
- Review the test by cause, not only by topic.
- Select the two highest-frequency failure mechanisms.
- Repair each on a small clean task.
- Return after a delay with notes closed.
- Introduce a fresh representation or wording.
- Mix with nearby alternatives.
- Use a timed section only after the route is stable.
- Compare the next test by error family, not just total score.
Frequently asked questions
Does this mean school tests are unfair?
Not necessarily. A test may simply remove supports that were present during learning. That difference is useful evidence about what the learner can carry independently.
Should the tutor give fewer explanations?
No. Good explanation builds the initial model. The important addition is a later phase where the learner must reconstruct and use it without the explanation present.
Why can my child explain but not answer?
Explanation and execution overlap but are not identical. A learner may understand a concept verbally yet still struggle with method selection, symbolic execution or answer precision.
What if marks improve only in tuition tests?
Compare the conditions. Topic cues, recent practice, familiar formatting and tutor-created question style may make tuition tests easier to navigate even when they are academically challenging.
Where this page sits in Atlas V2.0
This symptom page routes through the five learner states under When Learning Slips. Use Finding the First Weak Link when several causes are plausible, and Atlas V2.0 for the wider ecosystem route.
Last updated: 26 September 2026.
