eduKate Sengkang

Small Group Tutorials

Here to help students catch up, keep up, and move ahead. Book a consultation here.

Chat on WhatsApp

How to Perform in the new G3 SEC Examinations | Learner’s Guide Vol 0100 | Science: Repeat Measurements and Drift — Is the Pattern Real or Is the Reference Moving?

G3 SEC Science · Learner’s Guide Vol 0100. Learn to check repeatability, accuracy, reference drift and the order of trials through original evidence exercises.

A student measures the same object three times and obtains 5.2, 5.2 and 5.2 units. That looks reassuring. But suppose the instrument was zeroed incorrectly and always reads 0.4 units too high. Repetition can produce consistent results without making them accurate. A second student obtains 5.1, 5.3 and 5.2 with a correctly checked instrument. Which result is better? You cannot decide from neatness alone.

G3 SEC Science learners need to distinguish a change in the system from a change in the way the system is measured. A temperature sensor may warm during use. A balance may drift away from zero. A sample may change while it is being measured. The order of trials may affect the later trials. Simply writing “repeat and find the mean” does not solve every one of these problems.

Learner’s Guide Vol 0100 shows you how to compare repeated measurements, recognise patterns of drift, use an appropriate control or reference check, plan fair sequences and explain what the evidence establishes. The examples draw on Physics, Chemistry and Biology but remain original classroom reasoning cases. Use only the parts relevant to your actual subject combination, and carry out real laboratory work only under school supervision.

Chapter index

1. Repeatability and accuracy answer different questions

Repeatability concerns how closely measurements agree when made repeatedly under specified similar conditions. Accuracy concerns closeness to an appropriate true or reference value. A measuring instrument can be highly repeatable but systematically wrong. It can also be accurate on average while individual readings vary. A series of closely grouped numbers does not prove the measuring system is correctly calibrated.

Consider a standard reference mass of 100.0 g. A balance reads 102.0, 102.0 and 102.1 g across three checks. The readings are close to one another, but they are around 2.0 g higher than the stated reference. Under this hypothetical setup, the difference suggests a systematic issue that repeating the same procedure alone will not remove. The correct first response is to check the instrument, zero and method rather than announce that the result is accurate because the readings agree.

Now imagine another instrument reads 99.8, 100.2 and 100.0 g for the same standard. Their mean is 100.0 g. The readings are more spread out than the first balance’s values, but they cluster around the supplied reference. That does not mean this instrument is perfect; it means the reference check supports a different conclusion from the first series.

When an exam asks why repeating readings is useful, you can reasonably say that repetition helps assess consistency and identify unexpected readings. But if a problem explicitly shows an instrument bias or drifting reference, you need a more targeted improvement. A repeated systematic error remains systematic.

One strong habit is to state what repetition can reveal and what it cannot. “The three readings are consistent” is a direct observation. “The instrument is accurate” needs an appropriate reference check. “The measurement is reliable in every context” is much too broad for three readings.

2. Confirm the official G3 Science route

The SEAB 2027 G3 school-candidate syllabus list includes combined Science K326 (Physics/Chemistry), K327 (Physics/Biology) and K328 (Chemistry/Biology). Pure Physics K323, Chemistry K324 and Biology K325 are separate syllabuses. Confirm the exact combination you study before assuming a topic or paper applies to you.

Under the combined Science K326–K328 syllabus, Paper 1 consists of multiple-choice questions and lasts 1 hour, carrying 40 marks weighted at 20%. Each relevant disciplinary structured-and-free-response paper lasts 1 hour 15 minutes, carries 65 marks and contributes 32.5%. Paper 5 is a practical test lasting 1 hour 30 minutes, carrying 30 marks and contributing 15%. Only the relevant two of Physics Paper 2, Chemistry Paper 3 and Biology Paper 4 are taken for the combination.

The syllabus lists observing, measuring, interpreting experimental results, evaluating methods and suggesting improvements among the assessed experimental skills. It does not say every practical task will ask students specifically to diagnose drift. Our workshop uses drift and repeated readings as a way to practise those published skills.

For the broad Science knowledge behind the examples, start with the Science Hub. For a fuller foundation in measurement uncertainty, use Volume 0072 on uncertainty and anomalies. That earlier volume owns general precision and anomalous values. This volume asks a narrower question: what if the measurement reference or system changes during a series of trials?

Do not memorise one improvement phrase and paste it into every practical response. The purpose of a suggested improvement is to address a specific threat to the investigation. A systematic drift problem needs a different remedy from ordinary small random variation.

3. What a moving baseline looks like

A baseline is an agreed reference condition or reading. If the baseline changes during the investigation, later measurements may not be comparable with earlier ones. Imagine a device that should read zero with no load. At the beginning it reads 0.0 units. After ten minutes it reads 0.3, and after twenty minutes 0.6, even with no load. This is a changing reference.

Suppose an unknown sample reads 5.0 units at the start and 5.5 after twenty minutes. A learner may conclude the sample’s measured property rose by 0.5. But the reference drifted by 0.6 over the same period. The observed change could be explained partly or entirely by the measurement system. The data do not isolate the sample’s true change.

In a simplified additive-bias model, one might subtract a measured zero offset from each contemporaneous reading. If the unknown at twenty minutes reads 5.5 and the zero offset then is 0.6, the corrected value is 4.9. That is only justified if the instrument’s error behaves as an additive offset under the supplied model. Do not assume every sensor error can be corrected by simple subtraction.

The proper improvement might be to check a reference at multiple times, recalibrate as appropriate or use a more stable measurement method. If the reference readings themselves fluctuate, they need interpretation too. A correction based on one uncertain zero reading may not establish a highly precise result.

Notice that a moving baseline is not the same as an experimental variable changing intentionally. An investigator might deliberately heat a sample and expect its temperature to rise. That is the target effect. A temperature probe that drifts on a constant reference is a different effect: it interferes with confidence about the target measurement.

A useful notebook records both the sample and the reference condition, with timestamps. The purpose is to keep each reading attached to the state of the apparatus when it was taken.

4. Physics case: repeated measurements of a constant object

A pupil uses an instrument to measure the length of a rigid object. A controlled class exercise assumes that the object’s length remains effectively constant during the measurement period. The recorded values are 15.02, 15.04, 15.03, 15.05 and 15.04 cm. The range is 15.05−15.02=0.03 cm. The mean is (15.02+15.04+15.03+15.05+15.04)/5=15.036 cm.

Whether to report that mean to three decimal places depends on the measurement method and meaningful precision. The arithmetic average is 15.036, but the instrument’s stated resolution and uncertainty should govern reporting. A final answer with more digits than the method can justify gives a false impression of precision.

Now imagine a second series, measured with the same instrument as it warms or shifts: 15.02, 15.07, 15.12, 15.17 and 15.22 cm. The values rise by about 0.05 cm per reading. The object has been specified as effectively constant. The trend invites a check of instrumental or procedural change. Averaging the five readings would produce a number, but it would hide the systematic progression.

You cannot simply declare the instrument at fault without considering all supplied conditions. Perhaps a new operator changed the alignment or a reference mark moved. A fair diagnostic test measures a suitable fixed reference before and after the run, maintains consistent technique and checks whether the reading progression persists.

This case shows why measurement order matters. If all early values are taken with one setting and all late values with another, a drift could masquerade as a treatment effect. Alternating properly controlled conditions may help separate time trends from condition differences, provided the procedure permits it and there are no carryover effects.

A good answer says what the readings do: “The measured length rises systematically across the sequence although the object is assumed constant. The result suggests a time-related measurement problem and warrants reference checks.” That is better than “the object is growing”.

5. Chemistry case: sample change versus instrumental change

In a Chemistry investigation, a balance may measure total mass, a thermometer may track temperature or a suitable device may record an observable property. The measured sample may genuinely change during a reaction or physical process. A rising or falling series is not automatically drift; it can be the phenomenon under study.

Consider a data set showing the mass of a vessel and contents as 150.0, 149.7, 149.5 and 149.4 g at regular times, under an arrangement where loss of material to the surroundings is possible. The decrease could reflect a real loss under the stated setup. To evaluate whether the change is trustworthy, you might also check a constant standard mass on the same balance at suitable times. If the standard also declines, instrument drift becomes a plausible contributor.

Suppose the standard reads 100.0, 99.9, 99.8 and 99.7 g at the same observation times. The systematic decrease in the standard reading is suspicious. But a simple correction of the sample result may still require careful assumptions about the relationship between the reference drift and sample measurement. The central conclusion is that the sample trend should not be credited solely to the chemistry without checking the measurement system.

Now suppose the standard remains at 100.0 g while the sample mass falls under a method designed to allow gas to escape. That supports the view that the mass change is associated with the sample setup rather than the balance’s zero shifting. It does not by itself identify the gas or establish every chemical explanation. Further context is needed.

A good scientific method includes relevant control measurements, consistent positioning and appropriate handling of materials. Reweighing the same partly reacted sample is not necessarily a fresh repeat of the original reaction. A genuine repetition may require a fresh comparable starting sample under supervised laboratory procedures.

This connects to Volume 0096 on recovery after restoring a setting. The apparatus may be reset while the chemical composition is not. Repeating a reading, repeating a trial and restoring an original sample state are three different things.

6. Biology case: changing sample and changing measurement time

Living systems often change over time. A pulse-rate reading may decrease during recovery; a plant’s measured response may change after illumination is altered; repeated physiological measurements may vary naturally. If you record different values at different times, first ask whether time is part of the phenomenon rather than a nuisance.

Imagine a hypothetical pulse-rate exercise with readings of 112, 100, 91 and 84 beats/min taken after activity at fixed intervals. This is a time series, not four repeats of a fixed condition. Averaging those values would conceal the recovery pattern. The appropriate analysis is to describe the decline and the times at which readings were recorded, subject to the method.

If the goal is to assess repeatability of the pulse measurement at one phase, take repeated readings under suitably comparable conditions and methods. Readings taken immediately after activity and several minutes into recovery do not meet that requirement. The subject changed during the series.

A plant experiment introduces similar issues. Suppose bubble counts are recorded for equal one-minute intervals, with a lamp setting held constant. The counts change gradually. This might reflect an evolving system, external conditions or measurement variation. A repeated reference or control condition and careful timing can help distinguish them, but a simple mean alone does not identify a mechanism.

Sampling time matters especially when an intervention produces a delayed response. In Volume 0092 on delayed observations, a reading at minute two did not prove the process began at minute two. Here the same care is applied to repeated readings: they must be attached to the state and time of the system.

These are reasoning examples, not personal health tests. Avoid unsupervised physical or laboratory experiments, and do not draw medical conclusions from illustrative classroom readings.

7. Repeat a reading or repeat the experiment?

Repeating a reading means measuring the same or a closely comparable state again. Repeating an experiment means recreating relevant starting conditions and the procedure so that the outcome can be compared across runs. Those are distinct operations. A question may ask for repeated readings of a fixed length or repeated trials of a controlled reaction.

Suppose the goal is to find the length of a table. Measure it several times using a consistent, appropriate ruler method. Here the table is not expected to change between readings. Repetition can help reveal inconsistent placement or reading of the scale. A systematic zero error, however, needs a separate check.

Suppose the goal is to compare how quickly two similar samples change under different temperatures. Taking many readings on one sample during a single trial creates a time series. Performing another trial from a genuinely comparable starting condition creates a repeat. The distinction matters when analysing reliability and fair comparison.

Sometimes a procedure damages or consumes a sample. You cannot meaningfully repeat the original experiment on the same sample without recognising its changed state. Use fresh comparable materials under school guidance if the design requires an independent repeat. The second run should not begin halfway through the first run’s material history.

Other procedures have carryover effects. A sample that was heated may remain warm when the second condition begins. A sensor may retain a temporary offset or need a settling period. Alternating the conditions without resetting the relevant variables can make later readings depend on earlier treatments.

Write “What is being repeated?” beside a practice question. The answer might be the measurement of a constant object, the observation of a changing object at successive times, or the entire procedure with a new comparable sample. This small classification prevents an otherwise attractive but irrelevant improvement.

8. Why trial order can produce a false treatment effect

Imagine comparing condition A and condition B. A student always measures A first and B second. Over time, the instrument gradually shifts upwards by 0.2 units. If B readings appear slightly higher, is B the cause? Not necessarily. The later measurement may reflect the same time-related drift that would have occurred without changing the condition.

A simple illustration has reference readings of 10.0 at the start, 10.2 halfway and 10.4 at the end. A is measured near the start; B near the end. A difference of 0.4 could be explained by the time trend under a suitable model. If the test does not control timing, a claimed difference between A and B may not be isolated.

One possible remedy is to interleave A and B measurements where the method permits, include reference checks, and repeat the comparison with attention to order. A sequence such as A–B–B–A can sometimes help expose a time trend, but it is not universally suitable. In a chemical reaction with irreversible change, the sample may not be reset between conditions, and a new comparable trial may be necessary.

Why not just randomise everything? Randomisation can be useful in suitable designs, but it does not remove all problems. It cannot undo carryover effects or restore consumed reactants. The method must consider what a previous treatment leaves behind.

An exam answer should connect the improvement to the stated weakness. “Alternate the order in comparable fresh trials to reduce confounding between condition and time” is a clear reasoned proposal when appropriate. “Do more experiments” is less specific and may simply reproduce the same bias.

This is related to Volume 0068 on controls and alternative explanations. That volume examines threats to a fair test generally; this one focuses on the way the sequence of measurement can become an unwanted changing variable.

9. An original four-column audit

Try organising results in four columns: time, sample reading, reference reading and note about conditions. Suppose a sensor measures a sample at 0, 5, 10 and 15 minutes. Its sample readings are 4.0, 4.2, 4.4 and 4.6 units. A stable standard is also checked and reads 2.0, 2.1, 2.2 and 2.3. Both series increase steadily.

Without the reference, a student might interpret the rising sample reading as a true increase of 0.6 units. The reference shows a time-related increase of 0.3 units in a quantity intended to be stable. Under an explicitly additive-bias model, a corrected sample series could be 4.0−0.0=4.0, 4.2−0.1=4.1, 4.4−0.2=4.2 and 4.6−0.3=4.3, where reference deviations from 2.0 are subtracted. That suggests a smaller sample increase of 0.3 rather than 0.6.

But the subtraction is justified only if the exercise states or supports an additive common offset. A real apparatus might show changing sensitivity, nonlinearity or other effects. A reference trend is a warning and a diagnostic aid, not an automatic correction formula to use in every science question.

What can we conclude directly? The sample reading increases, and the reference reading also increases even though the reference is supposed to be stable. Therefore at least part of the apparent sample trend may be associated with measurement drift. More information about the instrument is needed to determine the corrected sample trend confidently.

Notice how this answer separates observations, model assumptions and interpretation. That distinction is highly valuable in practical-paper evaluation questions. It avoids both blind confidence and reflexive rejection of the whole dataset.

A good notebook makes the timing of each reference check visible. If the standard was only checked at the beginning, later drift would remain unobserved. If it was checked at inconsistent intervals, comparison might be less precise. Design the audit around the question you need the data to answer.

10. What does an anomalous reading actually tell you?

Suppose five repeated readings under a fixed condition are 8.1, 8.2, 8.1, 10.6 and 8.2. The fourth is conspicuously different. You can call it an anomalous reading relative to the others, but do not automatically erase it. First examine the measurement procedure, apparatus, recorded units and any unusual event that might explain it.

Perhaps the sensor was temporarily disconnected or a value was copied incorrectly. Perhaps the system genuinely changed for a short period. Repeating an appropriate measurement and checking the method can help, but an anomaly may contain useful information. Deleting it solely because it spoils a tidy average would be poor practice.

Now consider a sequence of 8.1, 8.2, 8.3, 8.4 and 8.5. No individual reading stands out dramatically, but the series has a systematic upward trend. This could be drift. Simply applying an “ignore the outlier” strategy would miss the problem because there is no isolated outlier.

The pattern’s shape matters. Random scatter around a stable level looks different from a monotonic shift over time. Yet a small set of readings cannot prove a complete causal account of the trend. A reference check and suitable controls are needed to distinguish competing explanations.

If asked how an anomalous result affects the mean, you may compute the mean with and without it for comparison, but do not silently omit data unless the question or justified method requires it. Explain the decision. A model answer that erases unexpected values without reason teaches the wrong scientific habit.

This extends the general measurement uncertainty and anomalies guide, Volume 0072. The specific issue here is recognising a trend across the full sequence rather than focusing only on one odd point.

11. Twelve original cases to diagnose

A. A standard with reference value 50.0 units gives readings of 52.0, 52.0 and 52.1. Are the readings consistent? Do they demonstrate accuracy? Explain the distinction.

B. A fixed reference is read at 0, 10 and 20 minutes as 0.0, 0.3 and 0.6 units. What concern does this raise for sample readings taken at the same times?

C. A rigid object is measured as 8.01, 8.02, 8.03, 8.04 and 8.05 cm in sequence. What systematic pattern should prompt a check? Is one reading an isolated anomaly?

D. A changing reaction yields total product volumes 0, 9, 16 and 21 cm³ at equal times. A learner calls them four repeats of a constant condition. What is the error?

E. A sensor measures 5.0, 5.3 and 5.6 units for the same supposedly stable sample, while its reference reads 2.0, 2.3 and 2.6. Under an explicitly additive offset model, calculate corrected sample values using the first reference as baseline.

F. A student records one trial at room temperature before lunch and one at a warmer room temperature afterwards. The second treatment is also different. Can the result isolate the treatment effect? Suggest a repair.

G. Readings are 4.1, 4.0, 7.2, 4.2 and 4.1. Identify the unusual reading and describe a scientifically responsible next step.

H. Repeated measurements are tightly grouped but all consistently above a supplied standard. Should the learner praise repeatability, question accuracy, or both?

I. A pulse rate is measured immediately after activity, then two and four minutes later. Should the three values be treated as repeated measurements of a fixed physiological state? Explain.

J. An experimental method consumes a reactant, yet the learner starts a second trial using the partly reacted mixture. What has changed, and why might the second trial fail to replicate the first?

K. Two sensor readings appear to agree after rounding to the nearest unit. Can that alone establish that the underlying quantities are exactly equal?

L. Explain the difference between “repeat the reading”, “repeat the experiment” and “check a reference standard”, with one concrete purpose for each.

Before reading the answers, classify every question. Is the key issue precision, accuracy, baseline, order, sample history or time-dependent response? Accurate classification often matters more than the length of the reply.

12. Worked answers and reasoned repairs

A. The three readings are highly consistent: their range is 0.1 unit. But they differ from the supplied reference by about 2 units. That is evidence of good repeatability in this short check and questionable accuracy under the stated conditions. Repeating the same method without reference adjustment may preserve the bias.

B. The instrument’s zero or reference may be drifting upwards. Sample trends could partly reflect that movement. The investigator should examine reference checks, calibration and whether the error behaves as an offset before correcting or interpreting the sample data.

C. The readings rise steadily, so the issue is a time-related trend rather than one isolated outlier. Check instrument alignment, zero, temperature, operator technique or another relevant changing condition under the stated setup. The observations alone do not identify the cause.

D. The values are a cumulative time series, not repeated measurements of a fixed state. The amount has increased as the process develops. The correct analysis uses changes between observations and their time intervals, rather than averaging them as though they were equivalent replicates.

E. The reference offsets relative to its initial 2.0 are 0.0, 0.3 and 0.6. Subtracting them from sample readings gives 5.0, 5.0 and 5.0 units. This answer relies on the explicitly supplied additive-offset model; such a correction should not be assumed for arbitrary instruments.

F. No. Treatment and room temperature both changed, so their effects are confounded. A repair is to compare conditions at suitably controlled temperatures, with comparable time and procedural steps and suitable repeats where appropriate.

G. The 7.2 reading is conspicuously higher than its neighbours. Review the apparatus, units and procedure, and make a justified repeat measurement under comparable conditions if possible. Do not discard it solely because it is inconvenient.

H. Both. The tight grouping suggests repeatability, but consistent disagreement with a suitable standard calls accuracy into question. The strengths and weaknesses are not mutually exclusive.

I. No. The subject is recovering, so time changes the state being measured. Treat the values as a response over time. Repeats for consistency would need comparable phases and methods, not arbitrary points along recovery.

J. The remaining reactant quantities or concentrations may differ from the original starting mixture. The second run is not equivalent to a fresh repeat. Use suitable fresh comparable starting material under school laboratory guidance where the design requires it.

K. No. Two rounded readings can hide small differences beneath the resolution or reporting level. They agree to the stated precision, not necessarily in every physical detail.

L. Repeating a reading helps assess consistency of a state under similar conditions. Repeating an experiment tests whether a procedure produces comparable outcomes from comparable starting conditions. A reference-standard check evaluates whether the instrument or method is aligned with a known quantity. Each addresses a different question.

These answer explanations should guide correction rather than become sentences to memorise. If you made a mistake, record whether you confused time series with repeats, missed the reference, ignored trial order or assumed that precision proves accuracy.

13. How to write a strong practical evaluation

An effective evaluation begins with the purpose of the experiment. If the aim is to compare two treatments, the main concern is whether the treatment alone explains the observed difference. If the aim is to measure an absolute quantity, calibration, zero and instrument resolution may be central. If the aim is to observe change over time, consistent sampling times and response lag matter.

Next identify one concrete weakness. “The balance zero changed from 0.0 to 0.6 g during the run” is stronger than “human error occurred”. A named weakness allows a named improvement and a clear explanation of its value.

Then propose an improvement fitted to the mechanism. “Check and record the zero at regular intervals, and recalibrate when appropriate” addresses baseline drift. “Repeat the measurement with consistent ruler placement” addresses an inconsistent length-reading method. “Use a fresh comparable sample for each trial” addresses consumed materials. One universal suggestion cannot solve all three.

Finally, explain how the change improves the evidence. It may make readings more comparable, reveal systematic shifts, reduce a known bias or improve the ability to distinguish competing explanations. Avoid claiming that one improvement “guarantees accurate results”. Scientific methods reduce identified weaknesses; they rarely make uncertainty disappear entirely.

The official combined Science syllabus values observation, interpretation and evaluation. Clear answers use scientific meaning in ordinary language. A concise explanation with an appropriate variable, unit and purpose is usually more useful than a long list of fashionable laboratory terms.

When a question asks for a limitation, give a limitation that actually follows from the described setup. If an instrument has a documented response delay, address the delay. If trial order is fixed while conditions drift, address order. Unsupported comments about “human error” or “small sample size” may not be relevant.

14. Seven days of deliberate practice

Day 1: Distinguish accuracy from repeatability with simple reference values. Describe what the readings show before recommending an improvement.

Day 2: Use a fixed standard across time. Draw a time series of reference readings and identify whether it is stable, scattered or steadily moving. Record the units and sampling times.

Day 3: Compare a sample time series with a reference time series. Explain what can be calculated and which correction would require an explicit model. Do not assume every drift can be subtracted.

Day 4: Practise separating repeated readings from repeated experiments. Use a length measurement, a chemical reaction and a hypothetical biological response.

Day 5: Investigate order effects through word problems. Ask whether A always occurring before B could confound a claimed treatment effect. Propose a realistic design suited to the process.

Day 6: Write short practical evaluations. For each flaw, produce a three-part response: weakness, improvement and reason. Include one situation where repeating does not remove systematic error.

Day 7: Attempt Questions A–L without notes. Revisit the categories you missed, and then analyse a fresh table. Transfer is demonstrated when you can read a new setup without reciting a model phrase.

Parents can ask, “What did you repeat—and was it really the same condition?” Teachers can ask, “How would you know if the instrument changed?” Both questions reveal whether the student understands the purpose of measurement rather than merely copying a data table.

15. Keep the reference in view

The central lesson is that data are produced by a system and a method together. Repeated numbers may reflect the underlying phenomenon, the measuring instrument, the procedure or a combination of them. One of the strongest examination habits is to separate those possibilities before claiming a cause.

Continue with the Science Hub and the G3 SEC learner-guide route. For continuity, link this lesson back to Volume 0092 on delayed measurements, Volume 0096 on recovery and Volume 0072 on uncertainty. Each teaches a different reason that a measurement may mislead.

Official examination details were checked against the SEAB 2027 G3 syllabus index and combined Science K326/K327/K328 syllabus. The numerical cases are original teaching examples and not official exam questions.