Measurement Uncertainty in Chemistry: A Practical Guide

No measurement is exact

No measurement is exact. A burette marked every 0.1 mL forces you to estimate the digit between marks; a balance flickers in its last digit; even the steadiest hand reads a thermometer slightly differently each time. Measurement uncertainty is the honest accounting of these limits — the range of values within which the true value reasonably lies.

Uncertainty is not an admission of failure. It is what separates a professional result from a bare number: $25.0 \pm 0.1$ mL tells a reader exactly how much trust to place in the measurement, while a naked 25.0 mL pretends to a perfection no instrument can deliver.

Absolute vs relative uncertainty

Uncertainty comes in two forms. Absolute uncertainty has the same units as the measurement and states the margin directly: $25.0 \pm 0.1$ mL, or $2.50 \pm 0.01$ g. It answers how wide is the range.

Relative uncertainty expresses that margin as a fraction of the measured value, usually as a percent:

$$\text{relative uncertainty} = \frac{\Delta x}{x} \times 100\%$$

It answers how big is the margin compared to the measurement. This matters because the same absolute uncertainty means very different things at different scales: ±0.01 g on a 2.50 g sample is 0.4%, but ±0.01 g on a 0.10 g sample is 10%. Relative uncertainty is the fair way to compare the quality of different measurements.

Where uncertainty comes from

Uncertainty originates in a few predictable places:

• Instrument precision: the smallest scale division sets the floor. A common working rule is an uncertainty of plus or minus half the smallest division — so a ruler marked in millimeters gives readings like $12.3 \pm 0.05$ cm.
• Estimation between marks: the estimated digit in any analog reading carries the bulk of the uncertainty.
• Calibration: even a perfectly read instrument inherits the uncertainty of whatever standard it was calibrated against.
• Random variation: repeated readings of the same quantity scatter, and that scatter is itself a measure of uncertainty.
• Digital resolution: a digital display that reads to 0.01 g has an uncertainty of at least ±0.01 g in the last digit.

Good practice is to record the uncertainty alongside the value from the start, rather than trying to reconstruct it later.

Propagating uncertainty: addition and subtraction

Measured values rarely stand alone — they get added, subtracted, multiplied, and divided, and the uncertainty travels with them. The rules for propagation of uncertainty are straightforward:

For addition and subtraction, add the absolute uncertainties. If you mix $25.0 \pm 0.1$ mL with $10.0 \pm 0.1$ mL, the total is $35.0 \pm 0.2$ mL. The margins simply accumulate because each measurement contributes its own wiggle room to the sum.

Notice the trap: subtracting two nearly equal numbers can produce a tiny result with a large relative uncertainty. $(25.0 \pm 0.1) - (24.8 \pm 0.1) = 0.2 \pm 0.2$ mL — a 100% relative uncertainty. Whenever possible, measure small differences directly rather than computing them from large numbers.

Propagating uncertainty: multiplication and division

For multiplication and division, add the relative (percent) uncertainties. If density is mass divided by volume, the percent uncertainty of the density is the sum of the percent uncertainties of the mass and the volume.

Worked example: mass $= 5.00$ g ± 0.4%, volume $= 2.00$ mL ± 0.5%.

1. Density $= 5.00 / 2.00 = 2.50$ g/mL.
2. Percent uncertainty $= 0.4\% + 0.5\% = 0.9\%$.
3. Absolute uncertainty $= 0.009 \times 2.50 = 0.02$ g/mL.
4. Reported result: $2.50 \pm 0.02$ g/mL.

A more rigorous statistical treatment combines percent uncertainties as the square root of the sum of squares, giving a slightly smaller value — but simple addition is a safe upper bound and is standard in introductory work.

The weak-link rule

Propagation leads to a practical rule: a result is only as precise as its least precise measurement — the weak link dominates. If one volume in a calculation has 2% uncertainty and everything else has 0.1%, the answer cannot honestly claim better than about 2%.

This is the physical meaning behind the significant figure rules: the decimal places you keep must reflect the actual uncertainty, not the calculator's display. Propagating uncertainty formally and applying significant-figure rules are two routes to the same destination — an answer that does not overstate what the measurements support.

Reporting results honestly

A properly reported result has three parts: the value, the uncertainty, and the unit — $25.0 \pm 0.1$ mL. The uncertainty typically carries one significant figure (two if the leading digit is 1), and the value is rounded so its last digit sits in the same decimal place as the uncertainty.

Remember the deeper point: uncertainty quantifies honest doubt, and honest doubt is a feature of good science, not a flaw. A chemist who reports $0.0800 \pm 0.0002$ M has told you something far more valuable than a chemist who reports 0.08 M and says nothing — the first result can be checked, compared, and built upon.

Variables and units

Common mistakes

Real-world applications

• Analytical laboratories report every result with its uncertainty to document reliability

• Pharmaceutical manufacturers work within strict tolerance limits on every batch

• Environmental agencies compare measured pollutant levels against legal limits using uncertainty ranges

• Engineering specifications define acceptable measurement tolerances for parts and materials

• Published scientific research requires uncertainty estimates for results to be reproducible