Wait, What? A measurement can look incredibly precise and still be wrong.

Suppose a balance displays 12.347 g. The five digits look impressive. But if the balance has not been zeroed properly, that value may be consistently displaced from the quantity you wanted. More digits do not automatically create better science.

Measurement quality is about asking several different questions that students often collapse into one: How close is the measurement to the quantity of interest? How consistent are repeated readings? How finely can the instrument distinguish changes? And how much doubt should accompany the reported value?

Accuracy and precision are not synonyms

Accuracy concerns closeness to the value of the measurand. Precision concerns the agreement among independent results obtained under stated conditions. A set of repeated readings can cluster tightly and therefore be precise while still being displaced from the accepted or reference value by a systematic effect.

NIST warns explicitly against using accuracy and precision as interchangeable labels. Its measurement guidance also distinguishes repeatability and reproducibility from the broader concept of uncertainty. Read NIST Technical Note 1297 here.

Resolution: what can the instrument actually distinguish?

Resolution is the smallest change that can be meaningfully distinguished by the measuring system. A ruler marked every millimetre has a different resolution from a measuring tape marked every centimetre. A digital balance reading to 0.01 g has finer displayed resolution than one reading to 0.1 g.

But finer resolution is useful only if the rest of the measurement process supports it. Reading 0.01 °C from a sensor does not guarantee that the actual experimental temperature is controlled to 0.01 °C.

Uncertainty is not an admission of failure

Every real measurement has limitations. Scientific reporting becomes stronger, not weaker, when those limits are made explicit. At school level, uncertainty may be estimated from instrument resolution, repeated measurements or the spread of readings, depending on context. At deeper levels, uncertainty analysis becomes a disciplined way of combining multiple contributions to measurement doubt.

The important conceptual move is this: a measured number is not a magical exact truth. It is an estimate produced by an instrument, method, calibration state, observer, environment and model.

Random variation versus systematic effects

Random variation causes repeated measurements to scatter. Repeating and averaging can reduce its influence on the estimated mean. Systematic effects shift measurements in a consistent direction or according to a consistent pattern. Repeating the same biased procedure does not remove that bias.

Example: timing a pendulum by hand introduces reaction-time variation. Timing 20 oscillations rather than one can reduce the fractional effect of the start-stop reaction time. But if the measured pendulum length is always taken from the wrong reference point, repeating the timing does not correct that length error.

Percentage uncertainty: why size matters

Suppose a measuring cylinder gives a volume uncertainty of about ±1 cm³. Measuring 10 cm³ gives a much larger fractional uncertainty than measuring 100 cm³. In simple form:

percentage uncertainty ≈ absolute uncertainty ÷ measured value × 100%

For ±1 cm³ at 10 cm³, the fractional uncertainty is about 10%. At 100 cm³, it is about 1%. This is why good experimental design often involves choosing quantities large enough that instrument limitations do not dominate the result.

Significant figures should follow evidence

A calculator may display 3.746281995. That does not mean the experiment supports nine significant figures. Reported precision should reflect the quality of the underlying measurements and the purpose of the calculation.

False precision is especially dangerous because it makes a weak result look authoritative. Good scientific writing avoids claiming more resolution than the experiment actually produced.

Choosing better apparatus

Ask: what quantity am I measuring, what range do I expect, and what resolution do I need? A burette is useful when delivering variable liquid volumes accurately. A volumetric pipette is designed for a fixed volume. A measuring cylinder may be sufficient for rougher work. The “best” apparatus is therefore not the most expensive one; it is the one whose characteristics fit the scientific job.

Observation versus inference

“The thermometer reading rose from 22.0 °C to 27.5 °C” is an observation. “The reaction released energy to the surroundings” is an inference supported by the observation and a model of energy transfer. Keeping this distinction visible helps students avoid pretending that an interpretation was directly measured.

Secondary → JC progression

Secondary: read scales correctly; avoid parallax; choose appropriate apparatus; record consistent decimal places; repeat readings; calculate means; identify anomalous values cautiously; recognise instrument limitations.

JC: reason quantitatively about uncertainty; propagate limitations through derived quantities where appropriate; understand calibration and zero errors; compare model predictions with uncertainty ranges; recognise when uncertainty is dominated by method rather than instrument resolution.

Checkpoint: which result is better?

The accepted value of a quantity is 50.0 units. Group A obtains 52.1, 52.0, 52.1, 52.0. Group B obtains 49.6, 50.4, 49.8, 50.2.

• Which group is more precise?
• Which group appears more accurate relative to the accepted value?
• What might Group A’s pattern suggest?
• Would taking twenty more Group A readings necessarily solve the problem?

Answer key and WHY reasoning

Group A is more precise because its readings are tightly clustered. Group B is less tightly clustered but centres much closer to 50.0, so it appears more accurate relative to the reference. Group A’s consistent displacement suggests a possible systematic effect. Taking many more readings may narrow the estimate of the same displaced mean but will not automatically remove the systematic cause.

How to raise your measurement standard

• Before measuring, predict the expected scale of the quantity.
• Choose apparatus with a range and resolution appropriate to that scale.
• Check zero, calibration state and reading method.
• Record raw data before calculating.
• Repeat where repetition reveals useful information.
• Separate random scatter from systematic mechanisms.
• Report only the precision justified by the evidence.

This transforms practical work from “take a reading” into evaluate what the reading deserves to mean.

Authoritative next steps

• NIST: Guidelines for Evaluating and Expressing Measurement Uncertainty
• NIST reference copy of Technical Note 1297
• SEAB 2026 O-Level syllabus directory

Teaching Guide

For parents and teachers: stop asking only “Did you get the right answer?” Ask “What limits this measurement?”, “What would repeating it tell us?”, “What would repeating it not fix?”, and “How many digits does the evidence really support?” Students improve rapidly when measurement vocabulary is attached to concrete mechanisms instead of memorised definitions.