A measuring instrument produces almost identical results every time. The numbers look stable. The display shows a decimal place. The operator repeats the procedure and obtains practically the same answer.

Should we trust the measurement?

Not necessarily.

One of the most important distinctions in physics and metrology is the difference between how closely repeated measurements agree with one another and how closely a measured value approaches the value of the quantity being measured.

These are related questions, but they are not interchangeable.

In Video Lesson 27 from Levitin Language School, we investigate this distinction through a simple example involving a digital balance and a reference weight.

Video language: Turkish. This English-language article provides an independent explanation of the scientific concepts presented in the video.

A balance that gives almost the same answer every time

Imagine a reference weight with an assigned mass value of 100.0 grams.

We place it on a digital balance and repeat the measurement four times under the same conditions.

The balance displays:

MeasurementIndicated mass
1105.1 g
2105.0 g
3105.1 g
4104.9 g

The results are remarkably close to one another.

Their range is only 0.2 g, from 104.9 g to 105.1 g. Their arithmetic mean is 105.025 g.

At first glance, the balance appears consistent.

But the assigned reference value is 100.0 g.

The difference between the mean indication and the reference value is approximately 5.0 g.

This creates an important scientific question:

How can an instrument produce highly consistent results while remaining substantially displaced from a reference value?

To answer it, we must distinguish several properties of measurement.

Precision: how closely do repeated results agree?

In metrology, measurement precision describes the closeness of agreement between indications or measured quantity values obtained by replicate measurements under specified conditions.

Precision concerns the dispersion of the results.

If repeated readings form a narrow cluster, they indicate relatively high precision under the conditions being examined.

Our balance produces values between 104.9 g and 105.1 g. That narrow spread is evidence of consistent behaviour in this particular series.

However, precision alone does not establish whether the readings are close to the reference value.

An instrument can be very consistent and still produce biased results.

Precision and repeatability are not identical

Repeatability is a particular form of measurement precision.

It applies when measurements are performed under repeatability conditions, including the same measurement procedure, operators, measuring system, operating conditions and location, over a short period of time.

Precision is the broader concept. Measurements can also be assessed under intermediate precision or reproducibility conditions.

Therefore, a narrow spread in four readings does not prove that the instrument will behave equally consistently in every laboratory, at every temperature or after prolonged use.

Repeatability describes performance under specified conditions, not a universal guarantee of consistency.

Accuracy: are we measuring the intended quantity correctly?

Measurement accuracy concerns the closeness of agreement between a measured quantity value and a true quantity value.

This differs from precision.

Precision examines agreement among repeated results.

Accuracy concerns agreement with the quantity’s true value.

In practical metrology, the true value is often not known exactly. Suitable reference values, calibration procedures and uncertainty evaluations are therefore essential.

For our example, we have an assigned reference mass of 100.0 g.

The balance repeatedly indicates approximately 105.0 g.

That comparison reveals a substantial discrepancy.

The measurements agree closely with one another, but they do not agree closely with the reference.

This is why a claim such as “the instrument is accurate because it always shows the same number” is scientifically unjustified.

What about measurement trueness?

A further distinction is useful.

Measurement trueness concerns the closeness of agreement between the average of an infinite number of replicate measured quantity values and a reference quantity value.

It is associated with systematic measurement error.

Trueness and precision are different properties, and measurement accuracy should not be reduced to either one alone.

In our example, the repeated indications suggest a systematic effect, but four measurements are not sufficient to establish every metrological characteristic of the balance.

Systematic error: when repetition does not remove the problem

Measurement errors may include systematic and random components.

Random effects can contribute to differences between repeated results.

Systematic effects can produce a persistent displacement in a particular direction.

Consider a balance that consistently indicates approximately 5 g more than the assigned reference mass.

Repeating the same procedure many more times might provide additional information about the distribution of its readings.

It will not automatically eliminate the systematic effect.

Suppose we obtain one hundred measurements, all clustered around 105 g.

Their average may become statistically more stable.

But a stable average around 105 g is not evidence that the 100 g reference weight has somehow acquired five additional grams.

More data can improve an estimate of repeatability-related variation while leaving an uncorrected systematic effect in place.

Repeating a biased measurement does not automatically make it unbiased.

Why more decimal places do not guarantee better measurements

Digital instruments can create a misleading impression of certainty.

Imagine two displays:

  • Instrument A: 25.000000 °C
  • Instrument B: 25.1 °C

The first display contains more digits.

But does that prove the first instrument measures temperature more accurately?

No.

The number of displayed digits is connected to the instrument’s indication and resolution. It does not, by itself, establish measurement accuracy or uncertainty.

An instrument may distinguish very small changes in its displayed value while still being affected by calibration limitations, environmental conditions, drift or the measurement method.

The same principle applies to our balance.

Displaying 105.000 g instead of 105.0 g would not eliminate the discrepancy with the reference value.

Display resolution is not the same as measurement accuracy.

Calibration: discovering a relationship, not automatically repairing a device

What should we do when a measuring instrument appears to have a systematic deviation?

One important step may be calibration.

Calibration establishes a relationship, under specified conditions, between quantity values provided by measurement standards, together with their uncertainties, and the corresponding indications, together with their associated uncertainties. This information can then be used to obtain a measurement result from an indication.

In simpler language, calibration helps us understand how an instrument’s indications relate to suitable reference values.

However, calibration and adjustment are not the same operation.

Calibration characterises the instrument’s measurement behaviour.

Adjustment changes the measuring system to obtain prescribed indications.

A calibration certificate does not automatically mean that the instrument has been adjusted, repaired or made suitable for every possible application.

For example, a balance may behave differently across its measurement range. Temperature, vibration, positioning and other operating conditions can also matter.

A reliable measurement process must consider the intended use, not merely the existence of a calibration certificate.

Measurement uncertainty is not simply the size of an error

Another essential concept is measurement uncertainty.

Measurement uncertainty is a non-negative parameter characterising the dispersion of quantity values being attributed to a measurand, based on the information used.

The measurand is the quantity intended to be measured.

In ordinary language, uncertainty helps us evaluate how the possible values attributed to that quantity are distributed, given the measurement model and available evidence.

It does not simply mean “how wrong the instrument is”.

Measurement error and measurement uncertainty answer different questions.

A measurement error concerns the difference between a measured quantity value and a reference quantity value.

Measurement uncertainty characterises the dispersion of values attributed to the measurand.

Sources of uncertainty can include repeatability, reference standards, calibration information, instrument resolution, environmental influences and assumptions within the measurement model.

These contributions must be evaluated appropriately rather than combined through guesswork.

Why we cannot invent an uncertainty interval

Our balance displays:

105.1 g, 105.0 g, 105.1 g and 104.9 g.

We can calculate their mean and their observed dispersion.

But those four numbers alone do not provide a complete measurement uncertainty budget.

We would need additional information about the reference weight, the balance, the measurement procedure, environmental conditions and relevant sources of uncertainty.

Writing a result such as “105.0 ± 0.1 g” without defining and justifying the uncertainty would create an appearance of scientific rigour without the necessary evidence.

Even when an uncertainty interval is properly reported, its interpretation depends on how the uncertainty has been evaluated and expressed.

The ± symbol alone does not establish a particular confidence or coverage level.

From measurement to scientific judgement

A measurement result is not automatically a decision.

In engineering, manufacturing, scientific research and laboratory practice, a measured value may need to be compared with a requirement or tolerance.

But the uncertainty associated with the measurement can affect the reliability of that comparison.

Consider a component with a maximum permitted length of 100.10 mm.

A measured value of 100.09 mm appears to fall inside the limit.

However, whether the component should be accepted may depend on the uncertainty and the decision rule used for conformity assessment.

This does not mean that every borderline result must be rejected.

It means that the conditions governing the decision must be understood.

The correct scientific question is not merely:

“What number did the instrument display?”

It is also:

“What evidence supports the result, and what decision can that evidence justify?”

Four questions worth asking about any measurement

When evaluating a measurement, consider four questions.

1. How consistent are the repeated results?

Examine their dispersion and the conditions under which they were obtained.

2. How does the result compare with a suitable reference?

A narrow distribution does not guarantee agreement with the reference.

3. Could a systematic effect be present?

Repeated measurements may share the same source of bias.

4. How has measurement uncertainty been evaluated?

The credibility of a result depends on the measurement model, available information and relevant limitations.

These questions help distinguish numerical appearance from scientific evidence.

Why this matters when learning physics and mathematics

A student may know how to calculate an arithmetic mean and still misinterpret the result.

Another student may recognise the formula for standard deviation but fail to identify a systematic measurement effect.

A third may assume that a digital instrument displaying six decimal places must be more accurate than one displaying only two.

These are not merely calculation mistakes.

They reveal a gap between performing a mathematical operation and understanding what the operation tells us about the physical world.

Physics requires models, evidence and interpretation.

Mathematics provides tools for describing quantities and relationships.

Scientific thinking connects those tools to the conditions under which a conclusion is justified.

At Levitin Language School, individual online academic learning is organised around understanding concepts, applying methods and explaining decisions rather than memorising formulas without context.

Students can explore physics, mathematics and other academic subjects through Levitin Language School, according to their learning goals and the available teaching specialisations.

Continue the investigation

This video lesson is based on a more extensive Turkish-language research article in the Language Thinking Laboratory:

Çok hassas bir ölçüm neden yanlış olabilir? Doğruluk, kesinlik ve belirsizlik arasındaki fark

The research examines further questions involving calibration, tolerance, measurement resolution, traceability, uncertainty and engineering decisions.

The full video lesson is presented in Turkish. This English page offers an independent route into the same scientific problem for readers who prefer to study the concepts in English.

You can also explore the school’s Video Lessons collection for other examples of learning through explanation, comparison and practical reasoning.

Final thought: a convincing number is not necessarily a reliable conclusion

Return to our reference weight.

Its assigned mass is 100.0 g.

The balance repeatedly indicates values close to 105.0 g.

The measurements agree with one another.

But they differ substantially from the reference value.

That difference is the heart of the lesson.

Precision, repeatability, accuracy, trueness, systematic error and measurement uncertainty are connected concepts, but each describes a different aspect of measurement.

A reliable scientific conclusion requires more than a stable display.

The value of a measurement lies not only in the number it produces, but in the evidence that allows us to interpret that number.


Tymur Levitin

Founder and Director, Levitin Language School

© Tymur Levitin — Levitin Language School. All rights reserved.