G2 Science K223, K224 and K225 data reasoning becomes more accurate when the learner stops asking only, “What does the graph show?” and starts asking, “Which exact claim does this evidence support?” The same table, graph or investigation can support one conclusion strongly, another only weakly and a third not at all. Scientific examination performance therefore depends on matching evidence to the scope, direction and mechanism of the claim.
This sixty-first Learner’s Guide develops a claim–evidence mapping system for G2 Physics, Chemistry and Biology. It is distinct from Vol 0036 on evidence strength, Vol 0044 on model versus measurement and Vol 0048 on data transformations. The narrow question here is: given a specific dataset, which claims are licensed by it, which require extra assumptions and which exceed the evidence?
The current 2027 G2 Science syllabus for K223, K224 and K225 assesses Knowledge with Understanding, Handling Information and Solving Problems, and Experimental Skills and Investigations. It includes interpreting information, identifying patterns, making predictions, drawing conclusions, evaluating methods and applying known principles to unfamiliar information. Those are official assessment directions; the claim–evidence map below is an eduKateSengkang reasoning framework.
The claim–evidence map
- Write the claim in one sentence before copying any data.
- Identify the exact observation or measurement that bears on that claim.
- Decide whether the evidence supports description, comparison, association, cause, prediction or mechanism.
- Check the conditions, range, comparison and measurement quality that limit the conclusion.
- Remove any wording stronger than the evidence actually earns.
The map is deliberately simple because it has to survive examination time. In training, the learner can expand each step and compare alternative claims. In the paper, the same logic can run in a few seconds.
Lab 1 — Difference is not automatically cause
Claim: Condition B produced a higher measured value than Condition A, so the learner wants to write that changing the tested factor caused the increase.
Evidence test: The numerical difference supports a comparison only if the two conditions were measured in a comparable way; a causal claim also needs the manipulated factor to be isolated.
Boundary: If temperature, starting amount, duration or another relevant condition also changed, the evidence cannot identify one cause cleanly.
Exam move: In the answer, state the measured difference first, then add causal wording only if the method genuinely protects the comparison.
Lab 2 — Trend requires a pattern, not one pair
Claim: The learner wants to conclude that as X increases, Y increases because one high-X condition has a higher Y value than one low-X condition.
Evidence test: A trend claim is supported more strongly when several levels of X show a consistent directional pattern rather than one isolated contrast.
Boundary: Two points can establish a difference, but they provide weaker support for a general relationship across an entire range.
Exam move: Describe exactly what the available points show and reserve broad trend language for evidence that actually spans the changing variable.
Lab 3 — Direct proportion is stronger than ‘goes up’
Claim: A graph rises from left to right, and the learner writes that Y is directly proportional to X.
Evidence test: Direct proportion requires the relevant constant-ratio or graphical condition, not merely a positive direction of change.
Boundary: Many increasing relationships curve, level off or change slope while still remaining positive relationships.
Exam move: Use ‘increases with’ unless the data satisfy the stronger mathematical condition needed for proportionality.
Lab 4 — Inverse proportion is stronger than ‘goes down’
Claim: Y decreases as X increases, so the learner labels the relationship inverse proportion.
Evidence test: Inverse proportion requires a reciprocal relationship such as an appropriate constant product, not merely a downward trend.
Boundary: A negative relationship can follow many functional forms and may flatten, steepen or approach a threshold.
Exam move: Describe the observed decrease first; promote it to inverse proportion only when the evidence supports that model.
Lab 5 — Observation and interpretation are different layers
Claim: A thermometer reading rises, and the learner immediately writes that an exothermic process occurred.
Evidence test: The rise in temperature is the direct observation; identifying the energy-transfer process is an interpretation supported by scientific knowledge and the setup.
Boundary: The measurement alone does not name every mechanism that could produce heating unless the question and method narrow the alternatives.
Exam move: When the command is ‘state what is observed’, give the reading change; when it asks ‘explain’, connect that evidence to the correct mechanism.
Lab 6 — No detected difference is not no effect everywhere
Claim: Two conditions produce similar readings, so the learner writes that the tested factor has no effect.
Evidence test: The investigation shows no clear detected difference under the tested range, sample, method and measurement resolution.
Boundary: A smaller effect, a different range or a more sensitive instrument might produce a detectable difference elsewhere.
Exam move: Keep the conclusion scoped to the investigation unless the scientific principle and evidence justify a broader statement.
Lab 7 — An anomaly does not erase a relationship
Claim: One point lies away from an otherwise consistent trend, and the learner concludes that there is no pattern.
Evidence test: Most of the measurements may still support an overall relationship while the unusual point remains part of the evidence that requires evaluation.
Boundary: The anomaly could reflect random variation, a method issue or a real change in system behaviour; the data alone may not identify which.
Exam move: Describe the overall pattern and the deviation separately, then judge whether the anomaly weakens, modifies or overturns the conclusion.
Lab 8 — Repeated readings do not fix every weakness
Claim: A method repeats each measurement five times, so the learner concludes that the investigation is accurate and valid.
Evidence test: Repeats can reveal variation and improve confidence in a mean by reducing the influence of random fluctuation.
Boundary: Repeating the same biased measurement or confounded comparison does not remove systematic error or make the design valid.
Exam move: Match the improvement to the problem: use repeats for random variation, and use better calibration, controls or method design for other weaknesses.
Lab 9 — A mean can hide variability
Claim: Two groups have the same mean, so the learner states that their results are the same.
Evidence test: The mean supports a statement about central tendency, not necessarily spread, consistency or individual values.
Boundary: One group can cluster tightly while another is widely dispersed even when the averages match.
Exam move: Use range, repeated values or another spread measure when provided before claiming that the groups behave similarly.
Lab 10 — Percentages need their base
Claim: Group A has 60% success and Group B has 50%, so the learner says more individuals succeeded in Group A.
Evidence test: The percentages show proportions, not raw counts; the group sizes determine how many individuals those percentages represent.
Boundary: A smaller group can have the larger percentage while containing fewer successful individuals.
Exam move: Identify the denominator before turning percentage evidence into a claim about counts.
Lab 11 — Rates and totals answer different questions
Claim: Condition A has the faster rate, so the learner concludes that it produced the greater total amount.
Evidence test: Rate describes change per unit time or another denominator; total amount depends on both rate and the duration or extent of the process.
Boundary: A faster process running for a shorter time can finish with a smaller total than a slower process running longer.
Exam move: Use rate evidence for speed-of-change claims and final amounts for total-output claims unless the full relationship is available.
Lab 12 — Final value does not show amount of change
Claim: Two samples end at the same temperature, so the learner says they changed by the same amount.
Evidence test: Change depends on both starting and final values; identical endpoints can follow very different changes.
Boundary: Without the initial measurements, the size and direction of change may be unknown.
Exam move: Map initial → final for each condition before comparing how much each changed.
Lab 13 — Model prediction and measurement must stay separate
Claim: A theoretical model predicts a value of 10.0 and the experiment records 9.7, so the learner ‘corrects’ the data to the model.
Evidence test: The model provides a prediction; the instrument provides measurement. Their difference is itself information about variation, assumptions and method.
Boundary: Agreement is not required to be exact unless the model and measurement precision justify it.
Exam move: State the prediction and the measured result separately, then discuss how closely they agree using Vol 0044 logic.
Lab 14 — Missing data are not zero
Claim: A table cell is blank, so the learner inserts zero and uses it in an average or trend.
Evidence test: A blank can mean not measured, not available or not reported; none of those states is automatically a numerical zero.
Boundary: Treating missing information as zero can create a false trend, change a mean and support a conclusion the experiment never measured.
Exam move: Preserve the missing state unless the question explicitly defines the blank as zero.
Lab 15 — Zero can be meaningful evidence
Claim: A reading is exactly zero, and the learner ignores it as ‘nothing’.
Evidence test: Zero may represent no detected output, a threshold condition, balance or another scientifically meaningful state depending on the variable.
Boundary: Its interpretation still depends on instrument resolution and the physical or biological system.
Exam move: Treat zero as data and ask what a zero value means for this specific measurement.
Lab 16 — Below detection is not the same as absent
Claim: An instrument reports no detectable signal, and the learner writes that the substance or effect is completely absent.
Evidence test: The measurement supports the claim that the signal was below the instrument’s detection capability.
Boundary: A smaller true value may exist below the method’s resolution, so absolute absence is a stronger claim than the evidence earns.
Exam move: Use detection-aware wording when the question provides information about measurement limits.
Lab 17 — Sample size limits scope
Claim: One plant, one trial or a very small group behaves a certain way, and the learner generalises to all organisms or all situations.
Evidence test: The sample supports a statement about the measured cases and may suggest a broader pattern.
Boundary: Broad population claims require adequate representation and repeated evidence; small samples are more vulnerable to unusual cases.
Exam move: Match the conclusion to the sample actually tested unless the question supplies a justified basis for wider transfer.
Lab 18 — Sample selection limits scope too
Claim: A selected group shows a pattern, and the learner assumes the same pattern applies to the whole population.
Evidence test: The data describe the sampled group reliably only if the measurement itself is sound.
Boundary: If the sample is unusual, biased or restricted to one condition, it may not represent cases outside that selection.
Exam move: Use any sampling information in the stem before generalising.
Lab 19 — Comparable methods are part of comparable evidence
Claim: Two conditions produce different readings, but they were measured with different instruments or procedures.
Evidence test: The values can be compared confidently only when the measurement meaning and method are sufficiently compatible.
Boundary: A method change can create an apparent difference even when the underlying system has not changed.
Exam move: Before attributing the difference to the scientific variable, check whether the measurement process itself changed.
Lab 20 — Controlled variables protect causal interpretation
Claim: The manipulated factor changes and the outcome changes, but several other relevant conditions were not controlled.
Evidence test: The result shows that the conditions differ and the measured outcome differs.
Boundary: Without control of plausible competing factors, the experiment may not isolate which change produced the outcome.
Exam move: State the observed difference confidently, but calibrate causal language to the quality of control.
Lab 21 — The independent variable is not enough by itself
Claim: The learner identifies the independent variable and immediately writes a causal conclusion without checking the outcome.
Evidence test: A manipulated factor creates the possibility of causal inference only when the dependent measurement actually responds and the design is valid.
Boundary: If the outcome does not change, or the method is confounded, the intended causal hypothesis is not supported by the result.
Exam move: Map changed variable → measured response before writing cause language.
Lab 22 — The dependent variable is not the original cause
Claim: The measured outcome changes strongly, so the learner treats it as the factor that caused the experiment’s condition.
Evidence test: The dependent variable is the response being measured; its change is evidence about what happened after the manipulation.
Boundary: Reversing cause and response changes the logic of the experiment.
Exam move: Reconstruct the design in words: what was changed, what was measured, and what remained comparable.
Lab 23 — One number rarely proves a trend
Claim: A single high value is used to justify a statement about how the system changes across the whole range.
Evidence test: One reading supports the state of the system at that condition.
Boundary: A trend requires evidence across multiple levels or times; one point cannot show the shape of a relationship.
Exam move: Use state language for individual readings and trend language only when the dataset contains a pattern.
Lab 24 — Extrapolation is not observation
Claim: The graph ends at X = 10, and the learner states what happens at X = 30 as though it was measured.
Evidence test: The measured range can support a model or trend used to predict beyond the data.
Boundary: The system may change regime, saturate or encounter a new constraint outside the measured range.
Exam move: Label beyond-range conclusions as predictions and use scientific principles to judge whether continued behaviour is plausible.
Lab 25 — Saturation changes the claim
Claim: Y rises with X at first, then levels off, yet the learner writes that increasing X always increases Y.
Evidence test: The data support an increase over the early range and a plateau over the later range.
Boundary: The relationship is not uniform across all tested values; a limiting factor or maximum may become relevant.
Exam move: Describe both phases and, if asked, use the scientific model to explain why the trend changes.
Lab 26 — Threshold behaviour changes the claim
Claim: Values remain near zero until one condition, then rise sharply, and the learner fits a simple steady trend.
Evidence test: The data support little change below the threshold and a different response above it.
Boundary: A single linear statement hides the regime change and can produce poor predictions.
Exam move: Name the threshold or change in behaviour when it is visible and scientifically meaningful.
Lab 27 — Conflicting datasets require evaluation, not cherry-picking
Claim: One dataset supports a strong effect while another, measured differently, shows a weaker effect; the learner cites only the preferred dataset.
Evidence test: Both datasets are evidence, but their methods, ranges, precision and controls determine how much weight each should receive.
Boundary: Choosing only supportive data can exaggerate certainty and hide methodological differences.
Exam move: Compare method quality and conditions before deciding whether the evidence genuinely conflicts or answers different questions.
Lab 28 — Correct mechanism cannot rescue irrelevant evidence
Claim: The learner writes a scientifically correct explanation but cites measurements from a different condition or variable.
Evidence test: Theory explains only when it is connected to evidence that actually bears on the claim.
Boundary: A correct textbook mechanism paired with irrelevant data remains an unsupported answer to the live question.
Exam move: Use evidence first, then mechanism. Make the line between the supplied data and the scientific explanation explicit.
Lab 29 — Relevant evidence does not require data dumping
Claim: The learner copies four or five numbers into a structured response because more data feel more convincing.
Evidence test: Often one comparison, one trend statement or one carefully chosen pair of values establishes the required pattern.
Boundary: Excess numbers can obscure which relationship actually supports the claim and waste time.
Exam move: Choose the smallest evidence set that proves direction, contrast or magnitude clearly.
Lab 30 — Alternative mechanisms need discriminating evidence
Claim: Two explanations are both consistent with the observed outcome, and the learner chooses the more familiar one.
Evidence test: If the mechanisms make different predictions under another condition, that condition can discriminate between them.
Boundary: When both mechanisms predict the same current result, the present dataset is insufficient to choose uniquely.
Exam move: Ask what additional observation would differ between the explanations instead of selecting by familiarity.
Lab 31 — Predictions inherit uncertainty
Claim: A noisy dataset produces a confident single prediction with no acknowledgement of variability or range.
Evidence test: The pattern may support a likely direction or approximate value.
Boundary: Sparse, variable or narrow data do not justify perfect certainty, especially outside the observed range.
Exam move: Use prediction language proportional to the evidence and the question’s expectations.
Lab 32 — MCQ options differ by evidence strength
Claim: Two options are scientifically possible, but one says the data ‘prove’ a cause while the other says they ‘show a relationship’.
Evidence test: If the design is observational or insufficiently controlled, the relationship statement fits the evidence better.
Boundary: The stronger causal option may be scientifically plausible but evidentially unsupported.
Exam move: Use claim strength as an elimination tool, extending Vol 0056.
Lab 33 — Structured responses need a visible evidence chain
Claim: The learner gives the correct conclusion but the answer does not show how the data support it.
Evidence test: A strong response names the relevant pattern or comparison and then connects it to the scientific principle.
Boundary: Without the evidence link, the answer can look memorised even when the conclusion happens to be correct.
Exam move: Use the compact order claim → evidence → reasoning, adding a limitation only where the question needs it.
Lab 34 — Physics: force and acceleration
Claim: For the same mass, larger applied net force is accompanied by greater measured acceleration, and the learner wants to generalise to all moving objects.
Evidence test: The controlled comparison supports the force–acceleration relationship under the tested conditions.
Boundary: It does not establish that every moving object has a net force in its direction of motion; constant velocity is a different state.
Exam move: Keep the evidence claim tied to acceleration and use the physical model for the mechanism.
Lab 35 — Physics: current and resistance
Claim: Current decreases when resistance is increased under the relevant controlled electrical conditions, and the learner attributes the change to an unmeasured battery problem.
Evidence test: The measured quantities support a relationship between resistance and current in the tested circuit.
Boundary: Inventing an alternative cause not supported by the setup weakens the answer even if that cause is possible in real life.
Exam move: Use the circuit evidence and syllabus relationship before introducing explanations outside the method.
Lab 36 — Physics: heating
Claim: One material’s temperature rises faster during comparable heating, and the learner says it always heats faster in every situation.
Evidence test: The data support a difference in thermal response for the tested masses, heating conditions and time range.
Boundary: Changing mass, power, starting temperature or geometry may change the comparison.
Exam move: State the measured result under the tested conditions and add the appropriate thermal explanation without universalising.
Lab 37 — Chemistry: rate versus yield
Claim: Two reactions reach the same final gas volume, but one reaches it sooner, and the learner says it produced more gas.
Evidence test: The time profile supports a faster rate; the same final volume supports similar measured final amount under those conditions.
Boundary: Rate and final yield are different claims even when they come from the same graph.
Exam move: Use slope or time evidence for rate and endpoint evidence for final amount.
Lab 38 — Chemistry: qualitative analysis
Claim: A test produces a characteristic observation, and the learner writes the inferred substance when the question asks what was seen.
Evidence test: The colour, precipitate, flame or gas-test behaviour is the observation; identification is the inference made from the valid test.
Boundary: Mixing observation and inference can lose marks even when the chemistry is known.
Exam move: Match the evidence layer to the command word before answering.
Lab 39 — Chemistry: concentration and rate
Claim: Higher concentration gives a faster measured reaction, and the learner writes that the table directly shows more frequent successful collisions.
Evidence test: The table supports the rate difference across concentrations.
Boundary: Collision frequency is a theoretical mechanism, not a directly measured column unless the method specifically measures it.
Exam move: Cite the rate evidence, then use collision theory to explain the pattern.
Lab 40 — Biology: enzyme response
Claim: Enzyme activity changes across temperature values, and the learner extends the exact same trend far outside the measured range.
Evidence test: The graph supports the relationship within the tested temperatures.
Boundary: At other temperatures the enzyme may behave differently, including plateau or loss of activity, depending on the biological system.
Exam move: Distinguish observed range from prediction and support the mechanism with appropriate enzyme knowledge.
Lab 41 — Biology: population change
Claim: A population declines after one environmental factor changes, and the learner assigns that factor as the sole cause.
Evidence test: The data show population change alongside the altered condition and may support causation if the design controls other relevant influences.
Boundary: Food availability, predation, disease or habitat could provide competing explanations in an uncontrolled setting.
Exam move: Use the design to decide whether the correct answer is association, causal effect or a more cautious conclusion.
Lab 42 — Biology: physiological measurement
Claim: Breathing rate rises during an activity, and the learner writes a complete internal mechanism as though every step was measured.
Evidence test: The recorded breathing rate supports a physiological response under the activity condition.
Boundary: Oxygen demand, respiration and transport mechanisms are scientific explanations inferred from system knowledge rather than all being directly observed.
Exam move: Separate measured response from explanatory mechanism while connecting them clearly.
Lab 43 — Unfamiliar context, familiar evidence logic
Claim: The question uses an unfamiliar organism, device or industrial process, and the learner assumes the conclusion requires outside knowledge.
Evidence test: The stem still provides variables, measurements, conditions and patterns that can be interpreted with syllabus principles.
Boundary: Surface novelty does not remove the need to respect evidence scope or command words.
Exam move: Strip the context down to claim, variable, evidence and relationship before deciding which scientific concept applies.
Lab 44 — Method change can create apparent scientific change
Claim: A second investigation uses a different apparatus and reports a different value, and the learner attributes the difference entirely to the scientific variable.
Evidence test: The observed values genuinely differ, but the changed measurement method may alter what or how accurately the system is measured.
Boundary: Without method equivalence, the comparison may mix system change with measurement change.
Exam move: Ask whether the method itself changes the meaning of the evidence before writing a causal conclusion.
Lab 45 — Control group absence limits attribution
Claim: A treatment group changes over time, and the learner states that the treatment caused the change even though no control comparison exists.
Evidence test: The data show change in the treated group.
Boundary: Background time effects, maturation or other factors could also explain the change without a comparable untreated condition.
Exam move: State the observed change and recognise that stronger causal attribution would require a suitable control or alternative evidence.
Lab 46 — Repeatability plus difference supports two claims
Claim: Repeated measurements are consistent within each condition and the condition means differ, and the learner blends these into one vague statement.
Evidence test: Within-condition consistency supports repeatability; between-condition difference supports comparison.
Boundary: Neither automatically proves the theoretical mechanism or absolute accuracy.
Exam move: State the two evidential achievements separately before adding any causal or mechanism claim.
Lab 47 — Claim ladders calibrate wording
Claim: The learner struggles to choose between ‘shows’, ‘suggests’, ’causes’ and ‘proves’.
Evidence test: Each word implies a different evidential burden: observation, association, causal attribution or very strong exclusion of alternatives.
Boundary: No universal vocabulary rule replaces reading the design; the same word can be appropriate in one experiment and too strong in another.
Exam move: Practise writing four versions of one conclusion and choose the strongest version the evidence genuinely earns.
Lab 48 — Evidence-swapping exposes ownership
Claim: A learner cites a correct trend from one graph to answer a question about another graph with similar variables.
Evidence test: Each dataset has its own conditions, range and method, so relevance depends on ownership as well as truth.
Boundary: Scientific correctness from the wrong dataset does not support the live claim.
Exam move: Label datasets during practice and require every conclusion to point back to its own evidence source.
Lab 49 — Missing-evidence questions improve experimental thinking
Claim: The current data cannot establish a desired conclusion, and the learner stops at ‘cannot conclude’.
Evidence test: Recognising insufficiency is correct but can be extended by identifying what new measurement or control would close the gap.
Boundary: The proposed extra evidence must target the actual uncertainty rather than simply add more data of the same uninformative type.
Exam move: Ask: what result would differ if the claim were true versus false? Design the next measurement around that distinction.
Lab 50 — Overclaim repair
Claim: A response says a factor ‘always causes’ an outcome after one limited investigation.
Evidence test: The experiment may strongly support an effect under the tested conditions.
Boundary: Always extends the claim to untested conditions and populations unless a general scientific principle and evidence justify that scope.
Exam move: Rewrite to the strongest bounded version the data support, keeping condition and range visible.
Lab 51 — Underclaim repair
Claim: A controlled dataset shows a strong consistent effect, but the learner writes only that there ‘might perhaps be a slight difference’.
Evidence test: Multiple comparable measurements support a clear directional difference.
Boundary: Excessive caution can make the answer vague and fail to state what the evidence actually establishes.
Exam move: Calibrate upward as well as downward: use direct language when the design and data genuinely support it.
Lab 52 — Diagnose the first reasoning failure
Claim: A conclusion is wrong, but the learner does not know whether to revise the concept, data reading or examination technique.
Evidence test: The written route usually reveals the earliest failure: wrong variable, wrong dataset, invalid comparison, overclaimed cause, missing mechanism or ignored condition.
Boundary: Fixing only the final sentence can leave the upstream reasoning unchanged and ready to fail again.
Exam move: Use Vol 0045: repair the first broken link, then retest with new data.
Use the Science Hub when the map exposes a knowledge gap
Claim–evidence mapping cannot replace the scientific principle itself. If the learner can identify the relevant data but cannot explain a circuit, particle process, reaction mechanism, enzyme response or ecological relationship, return to the Science Hub. Repair the concept, then retest it with a different dataset so the answer is not carried by memory of the first example.
Use the PSLE bridge without shrinking the G2 task
The PSLE Learner’s Guide series develops evidence-before-explanation, controlled comparisons and bounded conclusions. Those habits transfer. G2 adds denser disciplinary models, quantitative evidence and unfamiliar information, so the learner must preserve the same evidence discipline while handling a more demanding scientific context.
Use Examination Craft under time
The Examination Craft hub develops pacing, checking and recovery. Full four-line claim–evidence labs belong mainly in practice. In the examination, compress the routine to claim → evidence → link, then mark genuine evidential uncertainty for targeted checking rather than expanding every answer into a research report.
Readiness criteria
- You state the claim before selecting evidence.
- You distinguish observation, comparison, association, causation and mechanism.
- You know when a design is not controlled enough for a causal claim.
- You preserve conditions, range, sample and evidence ownership.
- You can state what a dataset does not show.
- You avoid turning ordinary trends into proportionality without evidence.
- You use claim–evidence logic in both MCQ elimination and structured responses.
- You can strengthen an underclaim and weaken an overclaim appropriately.
Official-source discipline
For current K223–K225 assessment objectives and paper structure, use the official 2027 G2 Science syllabus and the current SEAB G2 school-candidate directory. If SEAB updates the syllabus, the current official document takes priority over this guide.
Final rule: make the evidence earn every word of the conclusion
Science answers become stronger when claims stop floating above the data. Every conclusion should have an evidence path: what was measured, what pattern or comparison appeared, what the design allows you to infer and what scientific mechanism explains the result where required.
Do not ask only whether a claim is possible. Ask whether this evidence, from this method, under these conditions, is enough for that claim. That is the discipline that turns data into scientific reasoning.
