Did you know? Two homemade filters can remove different amounts of visible cloudiness and yet have the same proportional effect. A class that compares only the final jars may declare a winner that never had a fair chance to begin with. The mistake is not necessarily in the arithmetic: it can be hidden in the starting water.
A fair school water-filter comparison tests different filter designs against the same starting material, under the same measurement and operating conditions, and records both the initial and final cloudiness. When the starting water differs, calculate the changes honestly but do not claim that the better-looking final sample proves one filter is superior. Re-test matched samples before making that comparison.
This is a supervised classroom investigation about the measurement of cloudiness, not a method for making drinking water. Do not drink any experimental water, even if it looks crystal clear. Official guidance distinguishes turbidity, a light-scattering measure of suspended particles, from microbiological safety: see the Australian Drinking Water Guidelines and the US Environmental Protection Agency’s drinking-water overview.
Why the clearer final jar can tell the wrong story
Suppose a student pours a very cloudy sample into Filter A and a moderately cloudy sample into Filter B. The final sample from B looks clearer. That observation alone is real, but the inference ‘B is a better filter’ is not established: B may simply have received easier water. A fair comparison needs equivalent input samples and consistent filter volume, media amount, flow conditions, container cleanliness and measurement technique.
In a school demonstration, one group may stir more sediment into its beaker or take water from the bottom of a settling jar while another takes water near the top. Both groups might build identical filters and still record different outcomes. That is a sampling problem, not a filter-design result. Before changing the apparatus, look at the point where the inputs diverged.
A complete numerical example with units and assumptions
Imagine two fictional filters tested with different starting water in a teaching scenario. A calibrated turbidity instrument reports values in nephelometric turbidity units (NTU). Assume the instrument is suitable and readings are taken consistently; these values are invented for calculation practice, not real laboratory measurements.
Filter A: before filtration, 80 NTU; after filtration, 24 NTU. The absolute change is 80 − 24 = 56 NTU. The proportional reduction is (80 − 24) ÷ 80 × 100 = 70%. Filter B: before, 40 NTU; after, 12 NTU. The change is 40 − 12 = 28 NTU; the proportional reduction is 28 ÷ 40 × 100 = 70%.
A had the larger absolute change, B had the clearer final sample, and both had the same calculated percentage decrease. Which filter is better? These observations do not decide. They describe performance on different inputs, not a controlled head-to-head test. Even matching percentage reductions from different mixtures cannot establish that designs behave identically under all conditions.
To check the arithmetic independently, notice that 30% of 80 is 24, and 30% of 40 is 12. That confirms both samples retained 30% of their initial numerical turbidity values under the illustrative calculation. It does not turn turbidity into the mass of sediment: NTU is an optical reading, and its relationship to particle mass depends on the particles and instrument.
The matched-sample retest
Now imagine the class prepares one mixed batch of cloudy, non-potable test water, keeps it evenly mixed during portioning, and gives 100 mL of the same batch to each design. Both initial aliquots measure 60 NTU. Filter A ends at 18 NTU and Filter B at 9 NTU. A’s decrease is (60 − 18) ÷ 60 × 100 = 70%; B’s is (60 − 9) ÷ 60 × 100 = 85%. On this particular hypothetical matched trial, B produced the lower final turbidity and greater proportional reduction.
That observation is stronger evidence for this trial under these conditions, but it is still not a guarantee. Repeat independent runs using fresh, equivalently prepared samples and record how variable the results are. If one filter flows much more slowly, the comparison should also disclose the time or volume processed. Choose the performance criterion before looking at the results so the winner is not selected by whichever number happens to flatter it.
A learner diagnostic before the next experiment
Ask a child to answer without a formula sheet: ‘Why isn’t 12 NTU automatically proof of a superior filter when the other jar is 24 NTU?’ A sound explanation names the different starting conditions. If the pupil says ’12 is smaller than 24′, they can compare values but have not controlled the variable. If they calculate percentages correctly but conclude the materials are equally good, they are treating a single percentage as a causal test.
Then ask a second question: ‘If both filters begin at 60 NTU but one ends at 18 and the other at 9, what changed in the quality of the comparison?’ The answer is that the comparison now controls starting turbidity more closely, although other relevant conditions and repeated trials still matter.
Build the classroom procedure from guided to independent
First: define what success means
Choose a simple question: ‘Which of these two designs produces lower turbidity from equal portions of the same prepared water?’ Avoid replacing it with ‘Which makes safe drinking water?’ That is a different, much more demanding question requiring validated treatment and microbiological controls. Write the independent variable (filter design), dependent measure (resulting turbidity), and controlled factors (input batch, volume, container, timing, measurement conditions).
Second: diagnose the sampling setup
Use identical labelled containers and, with appropriate supervision, prepare one consistent water batch. Take two portions while the mixture is as uniform as practicable. Read the baseline value for each portion. If they differ meaningfully, stop and correct the portioning method instead of pretending the trial began fair. Follow school safety procedures for handling and disposal.
Third: run and document each trial
Record the filter design, amount and type of filter media, starting volume, before and after NTU readings if an instrument is available, elapsed time and unusual observations such as leaks or bypass. Label each run individually. Never erase an inconvenient result; explain a suspected mishap and repeat the run while preserving the original record.
Fourth: repeat, compare and test transfer
After showing the 80-to-24 and 40-to-12 example, offer the matched 60-to-18 and 60-to-9 problem without the model visible. Ask for absolute change, percentage reduction and a conclusion that explicitly says ‘in these trials’. Then change the baseline independently to 50 NTU and provide a final value of 15 NTU. The learner should obtain (50 − 15) ÷ 50 × 100 = 70% and explain that percentage reduction is still not the only performance measure.
A practical recording sheet
- Input check: source batch, portion volume (mL), initial reading (NTU or an explicitly described visual score).
- Filter record: materials, order, mass or layer thickness where measured, leaks or bypass, flow time.
- Output check: final reading using the same method and apparatus, with units.
- Arithmetic: initial minus final; percentage change only when the numerical scale and calculation are appropriate.
- Conclusion: what the specific data show, what they do not show, and whether repeated trials agree.
- Safety: experimental water never consumed; appropriate cleanup and equipment precautions supervised by an adult.
A photograph can help document appearance, but camera auto-exposure, lighting, jar shape and background colour can change how cloudy a sample appears. If a class has no turbidity meter, it may use a carefully defined qualitative appearance scale for comparison, clearly labelled as subjective. Do not calculate a scientifically meaningful ‘percentage turbidity removed’ from arbitrary categories such as ‘slightly cloudy’ and ‘very cloudy’.
Common misconceptions that make the result look more certain than it is
Misconception one: ‘Clearer means potable.’ No: clarity does not establish removal of microorganisms or dissolved contaminants. Misconception two: ‘The filter that removes more NTU must be better.’ Not when starting conditions differ. Misconception three: ‘A fair test means the jars look similar at the end.’ Fairness concerns the inputs and procedure, not identical outcomes. Misconception four: ‘One successful run proves the design.’ Single runs can be affected by chance variation, preparation differences or unnoticed leakage.
Students should also avoid describing turbidity reduction as removal of a known percentage of all bacteria, dissolved chemicals or total dirt mass. The instrument measures an optical effect under defined conditions. A student who can preserve this boundary is using scientific language more accurately than someone who reports a dramatic percentage without explaining what was measured.
What observable improvement looks like
Teachers can score four independent habits: identifying the controlling variable, naming the measurement units, interpreting both the baseline and final value, and limiting the conclusion to the actual evidence. Parents can ask the child to explain the same design with a fresh pair of numbers. A strong learner can reject a tempting but unsupported conclusion without needing the teacher to point out the unfair starting conditions.
An honest progress record might say: ‘First attempt: compared final values only. After explanation: identified different initial NTU readings. New trial: computed both percentage reductions independently. Later task: explained why equal baselines strengthen the comparison.’ These are invented rubric stages, not measured results from eduKate students or promises of examination outcomes.
Frequently asked questions
Should we always use percentage change?
Not always. Report the initial and final measurements and absolute change first. Percentage change is informative when the baseline is non-zero, the measurement scale is suitable and the interpretation is explained. It can be misleading if the instrument or optical scale is not comparable across different materials.
If both starting samples have the same NTU, is the comparison automatically fair?
No. Similar turbidity values are helpful but do not guarantee identical particle composition, volume, flow, filter loading or time. Control the other variables relevant to the research question.
Can children drink water after putting it through sand, cloth or paper?
No. This classroom experiment does not certify water as safe. Use only appropriately treated drinking water from a trusted supply for consumption, never experimental mixtures.
What if the result contradicts the prediction?
Record it honestly. Check the setup, repeat with a consistent procedure and revise the explanation in light of what was observed. A result that does not favour the preferred design is not automatically a failed experiment.
Where to deepen the learning
For the boundary between observation and conclusion, read how to tell a negative result from a failed experiment and how to measure fairly at different water depths. Continue through the eduKateSG Learning Hubs, or see the Clementi learning reference for the diagnosis–explanation–guided-practice–independent-check structure.
The real lesson is not which jar wins
An effective science lesson trains a child to ask what had to remain the same before the result could mean what we claim it means. Better filters are useful; better experimental reasoning is more transferable. The most important final answer includes the numbers, their units, the comparison conditions and the limits of what the evidence can prove.