Did you know that two places can have different temperatures at the same time without proving that sunlight alone caused the difference? A student records one warm corridor and one cooler corridor, makes an accurate bar chart, and announces that shade caused a three-degree reduction. That conclusion may sound sensible. But the corridors may also differ in ventilation, wall material, floor covering or the number of people passing through.
An observational study records what happens without deliberately assigning the condition being compared. A fair test, in common school Science usage, deliberately changes a chosen factor while trying to keep other relevant factors the same, and uses measurements to examine its effect. If a class merely compares two existing corridors, it has made an observation-based comparison; it has not isolated shade as the one tested cause.
The core aim is to teach a learner to use the right strength of claim for the way the evidence was collected. Good Science is not less exciting because we admit uncertainty. The best questions often appear precisely where an apparently obvious explanation has not yet been tested.
An invented temperature investigation on a school day
Imagine a wholly fictional school investigation. Corridor A is shaded by a roof. Corridor B receives direct sunlight for part of the morning. A class uses the same type of thermometer to record air temperature near each corridor at four clock times. These illustrative readings are not measurements from a real school. All values are in degrees Celsius (°C), with readings assumed to have been taken in the same order each round and after the instruments had time to settle.
- At 9:00 a.m., shaded Corridor A reads 29°C and sunlit Corridor B reads 32°C.
- At 10:00 a.m., Corridor A reads 30°C and Corridor B 33°C.
- At 11:00 a.m., Corridor A reads 29°C and Corridor B 32°C.
- At noon, Corridor A reads 30°C and Corridor B 33°C.
Both lists contain four values. Corridor A totals 29 + 30 + 29 + 30 = 118°C across four readings; its arithmetic mean is 118 ÷ 4 = 29.5°C. Corridor B totals 32 + 33 + 32 + 33 = 130°C; its mean is 130 ÷ 4 = 32.5°C. The difference between these mean readings is 32.5 − 29.5 = 3.0°C. At every matched clock time in this invented dataset, B is three degrees warmer.
Notice the units and wording. The sum of temperatures is an intermediate arithmetic step, not a meaningful total amount of heat. The useful summaries are the mean measured temperatures and their difference in degrees Celsius, not ‘three degrees of heat lost’ or a measured change in energy. Because the values repeat over times in a single hypothetical day, four entries also do not represent four independent studies.
What these results support—and what they do not
The observations support a descriptive sentence: ‘During these four fictional measurement times, Corridor B was 3.0°C warmer on average than Corridor A.’ They do not establish the sole cause of that difference. The observed corridors differ in more than the feature named in the conclusion. Both might receive different breezes, reflect heat differently, or sit close to different heat sources.
An evidence-to-conclusion mismatch arises when a student silently turns ‘was associated with’ into ’caused’. At times it is reasonable to propose a hypothesis such as, ‘Direct sunlight might contribute to the higher readings.’ That is explicitly a proposed explanation to investigate. It should not be written as if it were the only explanation already measured.
Scientific research faces the same challenge at larger scales. The AHRQ-supported research review on bias and confounding in observational studies describes how differing background factors can make causal conclusions harder to establish. That source concerns research on interventions, not classroom corridors, so we use it here only for the general idea of confounding rather than pretending it studied school weather.
Three features that distinguish a fair test from a comparison
- Manipulation: the investigator deliberately chooses which condition to vary rather than simply watching pre-existing differences.
- Control: important alternative influences are held as similar as practical or explicitly measured and addressed.
- Outcome: the same response variable is measured with a consistent procedure, units and enough repeated observations to judge whether the result is stable.
A fair test does not mean a perfectly controlled universe. It means the design aims to make the chosen factor the most credible explanation for an observed difference. The investigator should still state what could not be controlled. In a classroom, a safe and simple investigation often does this more clearly than a complicated apparatus with many hidden changes.
What would a stronger test of shade look like?
Suppose the class can use two otherwise similar empty demonstration spaces, identical temperature probes and a removable shade screen, with adult supervision and no unsafe heat sources. They might set up a plan in which location, measuring height, time, probe placement and nearby activity are kept as similar as practicable, while shade is intentionally varied. A stronger design could repeat and alternate which suitable position receives the shade, because permanent differences between positions could otherwise be mistaken for the shade effect.
Even that plan demands care: the screen itself may alter airflow; probe response time may differ; and the changing sun angle can shift heating during the observation. Those issues are not reasons to abandon the question. They become part of the method and limitations. If such conditions cannot be controlled in the school setting, the learners can continue with an honest observational comparison and refrain from a confident causal claim.
A classroom can also use a simulation to explore a model of light and temperature, but a simulation output is a result of its underlying rules rather than a direct measurement of a real corridor. The experimental question, model question and observational question must remain distinct.
A diagnostic question that reveals the central misconception
Ask a pupil: ‘Two potted plants grew to different heights. One stood near a window; one stood in a darker place. Did the class prove that light caused the difference?’ A correct answer is no, not from the description alone. Plants might differ in starting size, water, variety, pot, soil, temperature or care. The pupil should explain that these factors might affect growth and propose what the next investigation should control.
A response such as ‘Yes, because plants need light’ is an example of correct background knowledge being used to justify an inference the evidence did not actually isolate. Another weak answer is ‘No, we know nothing.’ We know there was an observed difference; we just do not yet know which cause explains it.
Progressive practice: three increasingly independent decisions
Stage one: classify the method
Statement A: ‘We measured two existing classrooms each day, one beside a busy road and one farther away.’ This is observational. Statement B: ‘We used otherwise identical model boxes and assigned one a different opening size while holding the remaining setup constant.’ This is a deliberate test of an assigned factor. Students must name what the investigator changed or did not change, not simply recognise familiar vocabulary.
Stage two: write a conclusion with the correct boundary
Give the class our corridor temperatures without the original conclusion. Invite two answers: one that states the observed mean difference, and one that proposes a possible cause. Students should write them separately. A good measurement sentence uses °C, four reading times, and the two mean values. A good hypothesis sentence uses ‘might’ or ‘could’ and identifies at least one alternative explanation that still needs checking.
Stage three: repair a flawed experiment independently
Present an invented plan: ‘Group A puts three seedlings outdoors in larger pots; Group B puts three indoors in smaller pots; after ten days, compare growth and conclude sunlight causes it.’ Ask learners to redesign it without a suggested checklist. They should notice that location, pot size and possibly watering or starting size differ. Their revision could use similar seedlings, matched pots and water, clear measurements of light exposure and growth over time, and repeated trials; they should still identify environmental influences that cannot be fully fixed.
Recording data so that later readers can audit the claim
- Write the question before measuring: are we describing two locations or testing a possible cause?
- Record where, when, how and with what each measurement was made.
- State the independent factor if one was deliberately manipulated, and the outcome measurement that is observed.
- Separate raw observations from calculations and interpretation.
- Identify possible confounding factors even when their effects were not quantified.
- Keep unexpected or inconvenient results unless a documented, justified correction is necessary.
- Use labels such as ‘hypothesis’, ‘observation’ and ‘conclusion’ accurately.
These records protect against a common classroom failure: beautiful graphs that cannot be traced back to a consistent procedure. A graph is only as informative as the measurements and definitions behind it.
What teachers, parents and learners can do next
Teachers can put two student conclusions beside one dataset and ask which claim is supported. Parents can practise the same skill with everyday observations—one room feels cooler, one route takes longer, or one plant looks healthier—without demanding a scientific experiment at home. Learners can improve by writing one factual observation, one possible explanation and one additional test for each scenario.
For an observable progress check, give a fresh scenario the following week with no word ‘fair test’ in the question. A student who can independently name the manipulated factor, list two realistic alternative influences and state what the data establish has moved beyond memorised definitions. A student who only repeats ‘keep everything the same’ still needs practice specifying exactly what should be comparable.
Limitations, precision and proportion
A temperature probe may respond slowly, warm in direct radiation or be positioned at different heights. A small set of convenient observations may not capture day-to-day variation. Correlation alone does not prove causation, but a carefully designed observational study can still yield important descriptions and inform better tests. Nor does deliberately changing one factor automatically guarantee causal certainty if the measurement process remains flawed.
Avoid implying that school investigations must use formal randomised trials. The scientific goal is proportionate reasoning: make a claim that matches the question, method, measurements and uncertainty. There are cases where manipulating a factor is impractical or inappropriate, and observation is exactly the right approach.
Frequently asked questions
Can an observational study be good Science?
Yes. It can reveal patterns, document conditions and generate questions worth testing. Its strength depends on accurate measurements and honest interpretation.
Does a fair test require every single condition to be identical?
No real investigation holds the whole world still. Identify the factors most relevant to the question, control or account for them where possible, and disclose limitations.
If two results differ by exactly 3°C, isn’t that proof?
It proves a difference in the stated illustrative readings, assuming they were accurately obtained. It does not by itself identify the cause of that difference.
What if a student cannot safely change the factor being studied?
Use observation, existing reliable data or a suitable model instead, and label what those methods can and cannot establish. Do not improvise unsafe experiments.
Further learning and a useful closing question
For another source of misleading results, see how to check for carryover between Science trials. For a wider route through careful investigations, visit the verified eduKate Education Hub. The immutable Secondary 1 Mathematics Clementi guide explains the habit of making hidden assumptions visible. Good education teaches a child to ask, ‘What did we actually change—and what did we merely notice?’