An anomalous result is a reading that does not fit the pattern of the others. The skill is to identify it, give a possible cause, and show the effect it has on the answer, rather than quietly deleting it.
This idea runs through cross-science data interpretation, whether the data comes from a titration, a pendulum or a plant experiment. All data here is invented.
How do I decide a result is anomalous?
Look for a reading that sits well outside the spread of its repeats, or a point that breaks a clear trend on a graph. Natural variation is normal: repeats rarely match exactly. An anomaly is a gap much larger than that natural spread.
Equally, an unexpected result is not automatically an error. If your repeats agree with each other, a surprising value may be real and worth reporting.
Step by step
- Check the repeats. Do most readings agree closely?
- Mark the outlier and state how far it differs from the others, with units.
- Suggest a plausible cause linked to the method, such as a late stop on a stopwatch or a misread scale.
- Repeat the measurement if you can. If the new reading matches the group, the odd one was probably an error.
- Report the mean without it and say so. Showing the mean with it as well is the clearest evidence.
Worked example
A student times 20 swings of a pendulum of length 40 cm, five times (invented data, seconds):
25.4, 25.6, 25.5, 28.9, 25.5
Spot: four readings lie within 0.2 s of each other. The fourth is 3.4 s above the typical value, so it is anomalous.
Cause: the stopwatch may have been stopped late or a swing may have been miscounted. This is a possible explanation, not proof.
Mean without the anomaly: (25.4 + 25.6 + 25.5 + 25.5) ÷ 4 = 102.0 ÷ 4 = 25.5 s.
Mean with it: (102.0 + 28.9) ÷ 5 = 130.9 ÷ 5 = 26.18 s, which is 26.2 s to 3 significant figures.
Report: “The reading of 28.9 s was left out because it differs from the others by far more than their spread. The mean of the other four is 25.5 s.” The time for 1 swing is then 25.5 ÷ 20 = 1.275 s.
The mistake to watch for
Mistaken answer: “I deleted the 28.9 s because it ruined my pattern.”
The reason given is about the student’s expectation, not about the data. A reader cannot check it.
The correction is to give an evidence-based reason, mention the deletion, and keep the original value visible in the table. The opposite slip is to average everything including the anomaly and never mention it.
Check yourself
1. Readings: 12.1, 12.3, 15.8, 12.2. Which is anomalous, and what is the mean without it?
Show answer
15.8 is anomalous. Mean = (12.1 + 12.3 + 12.2) ÷ 3 = 36.6 ÷ 3 = 12.2. With it, the mean would be 52.4 ÷ 4 = 13.1.
2. A point on a graph lies off the line, but its three repeats agree with it. Should it be removed?
Show answer
No. Agreeing repeats suggest the value is real. Mention it, and consider that the pattern may not be a simple straight line.
3. Why is “it did not match my prediction” a weak reason to discard a result?
Show answer
It describes the expectation, not a fault in the measurement. Data is allowed to disagree with a prediction. A valid reason links to the method, such as a misread scale.
Where this leads next
Revisit choosing the right graph to see how an anomaly looks on a plot, then continue to proportionality versus a general trend. The scientific investigation critic helps you word an honest evaluation.
If you tend to over-explain or under-explain odd results, our teachers can work through your own data in online one-to-one Combined Science tuition.