Validity, reliability and accuracyWhat each one means, how to tell them apart, and how to improve them
These three words turn up in almost every practical report and in most HSC science exams. Students often use them as if they mean the same thing. They don’t, and markers notice straight away. This guide explains each one using practicals students actually do at school.
Years 7 to 12
Physics, Chemistry, Biology, Science
Updated November 2026
The short answer
Validity
Did the experiment test what it was meant to test?
Only one thing was changed on purpose, the right thing was measured, and everything else that could affect the result was kept the same.
Reliability
Do you get the same result when you repeat it?
Repeat trials agree closely with each other. If they are all over the place, the results are not reliable.
Accuracy
How close is the result to the true value?
You can only judge this when you know the accepted value, such as g = 9.8 m/s² or the label on a bottle of vinegar.
An everyday example
Imagine someone who really weighs 60 kg stepping onto three different bathroom scales, five times each.
True weight: 60 kg
Scale A
63
63
63
63
63
Reliable, not accurate
Same reading every time, but 3 kg too heavy every time. Weighing again won’t fix it.
Scale B
57
63
60
62
58
Not reliable
The average is exactly 60 kg, but any single reading could be 3 kg out. You can’t trust one weigh-in.
Scale C
60
60
60
60
60
Reliable and accurate
Close to the true value, every time.
Now suppose the person uses Scale C, which works perfectly, to find out how much taller they have grown this year. The readings are still reliable and accurate, but the method is not valid. It measures the wrong thing, so it can’t answer the question.
Picture it as a target
Think of each dot as one result and the bullseye as the true value. Reliable results land close together. Accurate results land close to the centre. A method that isn’t valid is aiming at the wrong target altogether.
Reliable, not accurate
Not reliable
Reliable and accurate
Not valid: wrong target
Each dot is one result. The bullseye is the true value. In the last picture the results are close together and central, but on the wrong target. The small orange target is the one the aim was about.
You will also see the word precision, which describes how closely repeat measurements agree, so it sits very close to reliability. Our accuracy vs precision guide covers it in detail.
Validity
An experiment is valid when it actually tests the aim or hypothesis. In practice that means the result was caused by the variable you changed, and not by something else that crept in.
To judge validity, ask
Was only one variable changed on purpose (the independent variable)?
Were the other things that could affect the result named and kept the same (the controlled variables)?
Does the thing being measured (the dependent variable) really answer the aim?
Was the method able to measure it properly? For example, a reaction that finishes in two seconds can’t be timed well with a hand stopwatch.
How to improve it: identify any variable that wasn’t controlled and say exactly how you would keep it constant. “Use marble chips of the same size and the same total mass, 2.0 g, in every trial” earns marks. “Make it a fair test” doesn’t.
Reliability
Results are reliable when repeat trials agree with each other, allowing for the uncertainty in the measurement. If you did it again tomorrow, you would expect to get much the same answer.
To judge reliability, look at
How many trials were done. One result on its own can’t show reliability.
How close the repeats are. A small spread (range) means reliable results.
Whether any result is far from the others (an outlier), and whether there is a known reason for it.
How to improve it: repeat each measurement, usually at least three times, and use the mean. Then reduce whatever is causing the scatter. Timing 10 swings of a pendulum instead of one, for example, shrinks the effect of your reaction time on each result.
Some teachers and textbooks also talk about results being reproducible. That means a different person or group, following the same method, gets the same result. It is a stronger test than one student repeating their own trials.
Accuracy
A result is accurate when it is close to the true or accepted value. To judge accuracy you need something to compare against, usually a value from a data sheet or the known concentration of a standard solution.
To put a number on accuracy, work out the percentage error:
percentage error = | experimental value − accepted value |accepted value × 100%
Poor accuracy usually comes from a systematic error, something that pushes every result the same way. Common examples are a balance that isn’t zeroed, a thermometer that reads high, or heat escaping from a calorimeter.
How to improve it: find the systematic error and remove it at the source. Zero or calibrate the instrument, read scales at eye level, insulate the calorimeter and put a lid on it. Our systematic and random errors guide lists the common ones and their fixes.
Which one is it?
Each situation below mainly affects one of the three. Choose an answer to see if you’re right.
Investigating how length affects a pendulum’s period, a student also lets go from a bigger angle each time the string gets longer.
Validity. Two things changed at once, so you can’t tell whether the length or the angle changed the period. That is a validity problem.
Three titres in a titration were 24.10 mL, 25.60 mL and 23.40 mL.
Reliability. The repeats are up to 2.2 mL apart. Good titres agree to within about 0.10 mL, so these results are not reliable.
An electronic balance shows 0.15 g with nothing on it, and the student doesn’t press tare.
Accuracy. Every mass will be 0.15 g too high. That is a systematic error, which hurts accuracy. The readings could still agree perfectly with each other.
To see how light intensity affects photosynthesis, a student moves a hot lamp closer to the pondweed. The water warms up as the lamp gets closer.
Validity. Temperature also affects the rate of photosynthesis and wasn’t controlled. Any change in bubbling can’t be put down to light alone.
A class measures g three times and gets 9.2, 9.2 and 9.3 m/s². The accepted value is 9.8 m/s².
Accuracy. The results agree closely, so they are reliable. They are all about 0.6 m/s² below the true value, so the problem is accuracy.
A student times one swing of a pendulum by hand and gets 1.3 s, 1.6 s and 1.4 s.
Reliability. Reaction time makes each reading scatter. The fix is to time 10 swings and repeat, which improves reliability.
In an experiment on how acid concentration affects reaction rate, a different size of marble chip is used in each trial.
Validity. Chip size changes the surface area, which also changes the rate. It should have been a controlled variable.
A thermometer reads 2 °C higher than the true temperature for every measurement.
Accuracy. The same offset every time is a systematic error. The readings are consistent but not close to the true value.
Two practicals, worked through
Finding g with a pendulum
Physics, Years 11 and 12
Students measure the period T of a pendulum for five or six different lengths L, plot T² against L, and use the gradient to find g. The theory says T² = (4π² ÷ g) × L, so g = 4π² ÷ gradient.
Validity
Only the length is changed. The same bob and string are used, and the swing is kept small (under about 10°) because the formula is only accurate for small swings. Length is measured from the pivot to the centre of the bob.
Reliability
Time 10 full swings and divide by 10, then repeat three times for each length. Start and stop the timer as the bob passes the lowest point, where it moves fastest and the moment is easiest to judge.
Accuracy
A gradient of 4.10 s²/m gives g = 39.48 ÷ 4.10 = 9.63 m/s², which is 1.7% below 9.80 m/s². If every length was measured 2 cm short, the line would shift but its gradient would not, so g would be unaffected. That is one reason a graph beats a single calculation.
Titrating vinegar with sodium hydroxide
Chemistry, Year 12
Students find the concentration of ethanoic acid in vinegar by titrating it against a standard sodium hydroxide solution.
Validity
The colour change has to happen when the acid has just been used up, or you are measuring the wrong point. For a weak acid and a strong base that means phenolphthalein, not methyl orange.
Reliability
Do a rough titre first, then repeat until you have three concordant titres, usually within 0.10 mL of each other. Average only the concordant ones.
Accuracy
Standardise the sodium hydroxide first, because it absorbs carbon dioxide from the air and its concentration drifts. Rinse the burette with the sodium hydroxide and the pipette with vinegar, not water, or the solutions are diluted. Read the bottom of the meniscus at eye level, and add the base drop by drop near the end point so you don’t overshoot.
Markers reward answers that name the actual variable, error or step. Here is the same idea written two ways.
Weak
The experiment was valid because it was a fair test.
Strong
The experiment was valid because only the length of the pendulum was changed. The same bob and string were used for every trial and the release angle was kept under 10°, so changes in the period were caused by length alone.
Names the independent variable and each controlled variable.
Weak
To make it more reliable we could do it again.
Strong
Reliability could be improved by doing three trials at each length and timing 10 swings each time, which reduces the effect of reaction time on each result.
Says how many trials, what changes, and why it helps.
Weak
Our results were inaccurate because of human error.
Strong
Our value of g (9.63 m/s²) was 1.7% below the accepted 9.80 m/s². Some swings started at more than 10°, which makes the period slightly longer than the formula predicts, so g comes out too low.
Gives the size of the error, the cause and the direction it pushed the result.
Common mistakes
Blaming “human error”.
Instead: Name the actual error, such as reaction time or parallax, and say whether it is random or systematic.
Thinking more repeats make a result more accurate.
Instead: Repeats improve reliability and reduce random error. A systematic error stays exactly where it was.
Writing “it was a fair test” and stopping there.
Instead: List the controlled variables and how each one was kept the same.
Swapping reliable and accurate.
Instead: Reliable means the results agree with each other. Accurate means they agree with the true value.
Deleting an odd result because it “looked wrong”.
Instead: Only leave out an outlier if you know why it happened, such as the bob hitting the stand, and say so in the report.
Calling a whole experiment invalid over one small issue.
Instead: Explain how big the effect could be and which way it would push the results.
Exam-style questions
Try each one before opening the answer. The answers show the level of detail markers expect.
A student measured g three times and got 9.21, 9.24 and 9.19 m/s². Assess the reliability and accuracy of these results. 3 marks
Show a model answer
The results are reliable because they agree closely, with a range of only 0.05 m/s². The mean, 9.21 m/s², is about 6% below the accepted value of 9.80 m/s², so the results are not accurate. Because every result is low by a similar amount, a systematic error is the likely cause.
Distinguish between reliability and accuracy. Use an example in your answer. 2 marks
Show a model answer
Reliability is about whether repeat measurements agree with each other. Accuracy is about how close a measurement is to the true value. For example, a thermometer that reads 2 °C high gives reliable readings that are all inaccurate.
A student investigates how light intensity affects the rate of photosynthesis in pondweed by moving a lamp closer. Explain how the lamp could affect the validity of the investigation, and propose an improvement. 3 marks
Show a model answer
The lamp heats the water as it moves closer, and temperature also affects the rate of photosynthesis. The change in rate can’t be put down to light intensity alone, which reduces validity. Using an LED lamp, or placing a beaker of water between the lamp and the plant, and checking the water temperature with a thermometer, would keep temperature constant.
In an experiment to find the heat of combustion of ethanol, a student’s result was about 40% lower than the accepted value. Identify the most likely cause and explain how it affects the result. Suggest one improvement. 3 marks
Show a model answer
Most of the heat from the burner is lost to the air and the metal can instead of heating the water, and some ethanol burns incompletely (soot on the can). Because the water absorbs less heat than the fuel releases, the calculated heat of combustion is too low every time. This is a systematic error, so it reduces accuracy. Shielding the burner from draughts, putting a lid on the can and placing the flame close to the can would reduce the heat lost.
A student times one swing of a pendulum with a hand-held stopwatch. Explain two changes that would improve the reliability of the results. 4 marks
Show a model answer
Time 10 complete swings and divide by 10. The reaction-time error at the start and stop is then spread over 10 swings, so it has a much smaller effect on each period. Also repeat the timing three times at each length and calculate the mean, so that random variation between trials averages out and any outlier can be spotted.
For parents: a quick check at home
Questions students and parents ask
Can an experiment be reliable but not valid?
Yes. If an experiment measures the wrong thing, it can still give the same answer every time. Using an accurate set of scales to measure how much someone has grown is reliable and accurate, but not valid.
Can results be accurate but not reliable?
The average of scattered results can land close to the true value, partly by luck. You can’t trust it, though, because any single result could be far off. Scattered results point to large random errors that should be reduced.
Does repeating an experiment make it more accurate?
Only partly. Averaging repeats reduces the effect of random errors, so the mean can get closer to the true value. Repeating does nothing about a systematic error, because that error is the same in every trial.
What is a fair test?
A fair test is one where only one variable is changed and everything else is kept the same. It is the same idea as validity. In Years 11 and 12, students are expected to name the controlled variables rather than just say “fair test”.
Can one problem affect more than one of them?
Yes. In a heat of combustion experiment, heat escaping to the air makes every result too low, which affects accuracy. If more heat escapes with some fuels than others, the comparison between fuels is also less valid.
How many times should an experiment be repeated?
At least three trials is the usual school standard. Do more if the results vary a lot. Three trials that disagree are less useful than five that agree.