Accuracy and Precision Explained: Errors, Uncertainty and Anomalies
Learn the difference between accurate and precise results, how random and systematic errors change your readings, how to correct a zero error, and how to find a mean and an uncertainty and spot an anomaly. Also explains how AQA, Cambridge and the Punjab book use these words.
Published · 8 min read
On this page
- What does the target picture show?
- What is the difference between random and systematic errors?
- What is a zero error, and how do you correct one?
- What is resolution, and what is a least count?
- What do repeatable and reproducible mean?
- How do you find a mean and an uncertainty?
- Does precision mean the least count?
- How can you practise this?
- Common mistakes
- Key points
An accurate result is close to the true value. Precise results are close to each other: when you repeat a measurement, the readings bunch together. These are two separate ideas. Readings can agree closely with each other and still all be wrong. They can also be spread out but centred on the right answer.
Errors explain the difference. Random errors scatter readings up and down in an unpredictable way, so they spoil precision. Systematic errors move every reading the same way by the same amount, so they spoil accuracy but leave the readings bunched together. A zero error, where an instrument shows a value when it should show zero, is a classic systematic error.
Averaging several readings to get a mean makes random errors matter less. Repeating does nothing about a systematic error. You have to find its cause and correct for it.
What does the target picture show?
Picture four archers, each firing five arrows. The centre of the target stands for the true value. Each arrow is one reading.
| Where the arrows land | Precise? | Accurate? | Most likely cause |
|---|---|---|---|
| A tight group in the centre | Yes | Yes | Only small errors |
| A tight group off to one side | Yes | No | A systematic error |
| Scattered all around the centre | No | The mean is | Large random errors |
| Scattered, and off to one side | No | No | Both kinds of error |
The third row needs care. No single arrow is near the centre, but the average position of all five is. In the same way, the mean of scattered readings can still be close to the true value.
What is the difference between random and systematic errors?
Random errors change from one reading to the next. Examples include judging when to press a stopwatch, a draught nudging a balance, or a meter reading that flickers. A mean of many readings cuts their effect.
Systematic errors push every reading the same way. Examples include a zero error, measuring from the worn end of a ruler, or a thermometer that always reads 1 °C high. The mean of a hundred such readings is just as wrong as one reading.
Some errors can be either kind. Parallax error comes from reading a scale with your eye at an angle. If your eye is at a different height each time, the error is random. If you always look down on the scale from above, it is systematic. Put your eye level with the mark you are reading.
What is a zero error, and how do you correct one?
Before you measure, check what the instrument shows when it should show zero. Close the jaws of the callipers, hang nothing on the forcemeter, or empty the balance pan. Then use:
A zero error above zero is positive, so you subtract it. A zero error below zero is negative. Subtracting a negative number adds it on.
Worked example 1: a forcemeter and a thermometer
(a) A forcemeter shows 0.3 N with nothing attached. With a bag of rice hanging from it, it shows 4.7 N. What does the rice weigh?
Zero error: +0.3 N
Corrected reading: N
(b) A thermometer in melting ice shows −1 °C instead of 0 °C. In a classroom in Kuala Lumpur, it shows 24 °C. Assume the error is the same all along the scale. What is the real temperature?
Zero error: −1 °C
Corrected reading: °C
Answer: the rice weighs 4.4 N and the room is at 25 °C. Extra readings would never have revealed either error.
What is resolution, and what is a least count?
An instrument's resolution is the smallest change it can show. A millimetre ruler has a resolution of 1 mm. A digital thermometer showing 21.4 °C has a resolution of 0.1 °C. The Punjab textbook calls the same idea the least count.
Pick an instrument whose resolution is tiny compared with what you measure. A ruler suits a 30 cm book, but its 1 mm steps are a big share of a 2 cm coin. Vernier callipers, reading to 0.1 mm, suit the coin far better.
Resolution is not the whole story. A stopwatch may show hundredths of a second, but your reaction time adds random errors far bigger than 0.01 s.
What do repeatable and reproducible mean?
- Repeatable: the same person, with the same method and equipment, does the measurement again and gets similar results.
- Reproducible: another person, or a different method or set of equipment, gets similar results.
How do you find a mean and an uncertainty?
- Look for any anomaly: a reading that does not fit the pattern of the others. Check what went wrong, and leave it out or repeat it.
- Add up the other readings and divide by how many there are. This is the mean.
- Find the range: the largest reading minus the smallest.
- Take half the range as the uncertainty, and write the result as mean ± uncertainty.
Worked example 2: a toy car on a ramp
A student in Accra times a toy car rolling down a ramp five times: 1.84 s, 1.78 s, 1.82 s, 2.31 s and 1.80 s.
Step 1, the anomaly: 2.31 s is about half a second longer than the others, which all lie within 0.06 s of each other. Perhaps the car rubbed against the side. Leave it out, or do that run again.
Step 2, the mean: s
Step 3, the range: s
Step 4, the uncertainty: s
Answer: the time is 1.81 ± 0.03 s. Including 2.31 s would have given a mean of 1.91 s, longer than every one of the four good readings.
Does precision mean the least count?
There are two meanings, and you need to know both.
Meaning 1: how closely repeat readings agree. This is how AQA uses the word, and Cambridge's list of measurement terms gives it the same meaning. It is also where the Punjab book starts, with a target figure like the one above.
Meaning 2: how fine the instrument's scale is. The Punjab book goes on to say that the instrument sets the precision, so an instrument with a finer least count gives a more precise reading. Cambridge's advice on recording data uses this sense too, when it asks you to record readings to the precision of the instrument.
The Punjab book also links accuracy to the fractional uncertainty: the uncertainty divided by the measured value. For one reading, it takes the uncertainty as half the least count, so a millimetre scale gives ±0.05 cm.
Worked example 3: one ruler, two measurements
A student in Multan uses a tape measure with a least count of 0.1 cm, so each reading is ±0.05 cm. A desk measures 120.0 cm and a coin measures 2.3 cm. Which result is more accurate?
Desk: , or about 0.04%
Coin: , or about 2%
Answer: in the least-count sense, both readings are equally precise, because the ruler is the same. In the Punjab book's sense, the desk reading is far more accurate, because its uncertainty is a much smaller fraction of the value. It also has more significant figures (four against two). For the coin, use an instrument with a smaller least count.
In an AQA or Cambridge answer, use meaning 1 unless the question is clearly about the instrument's scale.
How can you practise this?
Open our vernier callipers simulation and choose Practice. Include zero errors is switched on at the start, so press Check zero before you read each object, then type in the corrected size. There is also a micrometer screw gauge simulation. To see accuracy and precision side by side, fire shots at a target, take repeat readings with faulty instruments and compare resolutions in the accuracy and precision simulation. For the scales themselves, see our guide to reading a vernier calliper and a micrometer.
Common mistakes
- Treating precise as accurate. Readings that agree closely can all share the same systematic error.
- Hoping repeats will fix a systematic error. Only finding and correcting the cause will.
- Adding a positive zero error. Subtract it. Only a negative zero error ends up being added.
- Keeping an anomaly in the mean. Check it first, then leave it out or repeat it.
- Using the whole range as the uncertainty. It is half the range.
- Quoting extra decimal places. Match the resolution of the instrument.
Key points
- Accurate means close to the true value. Precise means repeat readings agree closely.
- Random errors scatter readings: reduce their effect with a mean. Systematic errors shift every reading: find and correct them.
- Corrected reading = observed reading − zero error.
- Resolution, or least count, is the smallest step an instrument can show.
- Uncertainty = half the range of the repeat readings, after removing anomalies.
- The Punjab book also uses precision for the least count.