Did you know that a surprising number in a Science practical can be the most useful part of the lesson? Students often feel tempted to erase a result that looks wrong. But an unexpected reading might reveal a misread scale, a change in conditions or a real event worth investigating. The aim is to examine it, not tidy it away.
When a measurement disagrees with other readings, keep the original record, check the method and conditions, repeat an appropriate measurement or trial, and only then decide what the evidence supports. One unusual number is a reason to investigate, not automatic permission to delete data or announce a scientific discovery.
First ask: what exactly was measured?
A measurement is a recorded value of a defined quantity under particular conditions. That definition sounds formal, but it protects the learner from several common errors. “Length of flight” might mean distance from launch line to first landing, or distance to where a plane finally stops. If two trials use different endpoints, their numbers may not be comparable.
Before redoing the arithmetic, identify the instrument, its units, the starting mark, the endpoint, the person reading it, and anything that changed between trials. You cannot repair an unclear definition by simply repeating the same unclear procedure.
Worked example: three paper-plane trials
Imagine an illustrative classroom investigation comparing two folded paper-plane designs. Under a written testing procedure, the first three flights of design A land 6.1 m, 6.3 m and 6.0 m from the start. Three flights of design B land 6.2 m, 13.8 m and 6.1 m away. All readings are fictional teaching data, not outcomes from an actual class.
The 13.8 m reading stands out. It is more than twice either of B’s other readings. The arithmetic mean for A is (6.1 + 6.3 + 6.0) ÷ 3 = 6.13 m, to two decimal places. For B, the mean is (6.2 + 13.8 + 6.1) ÷ 3 = 8.70 m. If a pupil only compares means, B looks clearly ahead. But B’s median is 6.2 m; A’s median is 6.1 m. This does not prove which design performs better: three trials for each design are too little evidence for a confident general conclusion, especially with an unusual result.
What should a careful learner do? Keep the 13.8 m record. Verify that the start and landing marks were read consistently, ask whether one launch used a different height or received a gust of wind, and check whether the distance was written as 13.8 instead of 6.8 or another number. A transcription error is only one possibility; it must be confirmed before any correction.
Then arrange further fair trials, using the same measurement definition and comparable launch conditions. If the long flight is repeatable, it may be part of the genuine variation. If an identifiable procedure error affected that flight, mark it transparently and explain how it was treated. Do not silently erase it just because the pattern looks neater without it.
Three different diagnoses call for three different actions
- Recording error: a student measured 6.8 m correctly but typed 13.8 m into the table. Compare the original measurement note with the copied value; correct the copy and keep a trace of the correction.
- Measurement error: a tape began at the 1 m mark rather than zero, or a digital instrument was not zeroed. Fix the method, label the questionable readings and take new readings as appropriate.
- Real variation: one paper plane received a stronger gust or was launched differently. That is part of what happened. Improve control of conditions for the next investigation and discuss how variations affect the conclusion.
A fourth possibility is that an unexpected reading reveals a previously unrecognised effect. It deserves investigation, but one reading on its own cannot tell you what caused it. The important scientific habit is to separate what was observed from why it might have happened.
How to run a responsible recheck
Record before diagnosing. Copy the unusual value, units and test conditions exactly. Circle it or annotate it as ‘check required’. Do not overwrite a number from memory.
Inspect the measurement path. Walk backwards from the result to the instrument and procedure. Ask where zero was, whether the observer recorded first landing or final resting point, whether the tape was straight and whether the method changed during the trial.
Choose the appropriate repeat. Reread the same recorded measurement if it was transcribed incorrectly; remeasure the same object if the object is stable; repeat a full trial if conditions changed and a single measurement cannot be recreated. These are different activities, and a repeat of one cannot always substitute for another.
Keep comparison fair. For paper planes, use the same starting line and distance definition. Test both designs across comparable conditions, and record the order of trials so a later improvement or gust does not unfairly favour one design.
Write a limited conclusion. Instead of “Design B is best”, the student might write: “B had the larger mean in these six recorded flights because one B trial reached 13.8 m; the results vary and further comparable trials are needed before choosing a more reliable design.” This statement uses the observations without promising a broader result.
What an adult can ask without giving away the answer
Parents and teachers can point to the number and ask four calm questions: “What was measured?”, “What else changed?”, “Could we check that record?”, and “What conclusion would still be fair if we cannot explain it yet?” That sequence teaches investigation rather than compliance. Avoid supplying the explanation (“It was the wind”) before evidence exists. A student should be allowed to propose and reject several plausible explanations.
For younger learners, demonstrate with repeated ruler measurements of one notebook, then invite them to identify why starting at zero matters. For older learners, use a data table with known conditions and ask for a short evaluation of reliability, not just the mean. Match the sophistication to the schoolwork actually assigned; an open scientific investigation and an examination calculation do not always require the same response.
Common mistakes and how to repair them
- Deleting the outlier automatically: keep it first; explain any later exclusion with a documented reason.
- Repeating until the answer looks nice: decide the retest plan before seeing the later results where feasible.
- Calling one extra reading “proof”: more measurements can help, but repeated trials may share the same systematic error.
- Changing several conditions at once: a new outcome will be harder to explain. Record what changed and control relevant variables where possible.
- Confusing an average with typical performance: compare the raw observations, variation and a suitable summary before deciding what the numbers show.
- Claiming the expected result is always correct: predictions guide tests, but observations can challenge them.
A short independent check
Give the learner an invented temperature record from one stable cup: 24.1°C, 24.2°C and 34.1°C. Ask what they can safely conclude. A strong answer notices that 34.1°C is unusual relative to the others, checks the thermometer and timing, proposes a repeat measurement and refuses to change the value without evidence. The student may discuss possible environmental changes but should not assume them as fact.
Then offer a new scenario, such as a timed run or a mass measurement. Does the learner transfer the same record–inspect–repeat–explain logic without hints? Record which questions they ask themselves. Progress is visible when they begin checking units, endpoints and conditions before trying to ‘fix’ the number.
Important limits: precision, uncertainty and safety
Readings have limits. A ruler marked in millimetres cannot honestly justify unlimited decimal places. Changing the number of decimal places does not eliminate measurement uncertainty. Fair testing helps, but perfectly identical real-world conditions are rarely possible. Repeated readings are also not automatically independent trials: measuring the same flight distance three times is not three separate flights.
For practicals involving heat, electricity, glassware or chemicals, follow the teacher’s instructions and safety controls. Do not repeat a potentially unsafe test simply to remove a surprising result. The student can still record what happened and ask an educator how to investigate it safely.
Frequently asked questions
Should an outlier always be excluded from the mean? No. The decision depends on the purpose, method, measurement quality and justified analysis plan. Report exclusions transparently; never delete a result only because it weakens the preferred conclusion.
Does repeating a test guarantee the truth? No. Repetition can reveal inconsistency and improve evidence, but systematic errors can persist across every repeat.
Is a surprising observation a failed experiment? Not necessarily. It may reveal a useful limitation, a procedural issue or a real phenomenon. The quality of the investigation depends partly on how faithfully it is recorded and assessed.
What is the best first sentence in a conclusion? Begin with what the measured data actually show. Keep causes and wider claims proportionate to the evidence available.
The educational aim beyond one experiment
A student who can say “This result surprises me; I should check how it was measured before deciding what it means” is developing scientific judgement, not merely test technique. For complementary reasoning, read how predictions, observations and explanations differ and how to check transcription errors. The eduKate Education Hub provides a wider learning route, while the Clementi learning reference models the value of diagnosis, progressive practice and independent checking. A trustworthy conclusion should be earned by the evidence, not by a tidy-looking table.