ACCURACY, RELIABILITY AND VALIDITY – WHAT DO THEY REALLY MEAN?

Name: _____________________________________

Adapted from

 Curriculum Support,
Science, 2001 Vol. 6
No. 3
Low Accuracy High Accuracy High accuracy  www.curriculum-
High Reliability Low Reliability High Reliability press.co.uk: BioFact
Sheet 230

When doing experiments (first-hand investigations) it is important to understand the terms

errors, accuracy, reliability and validity. Understanding these terms allows a scientist to
confidently draw conclusions from the results of an investigation.

1. Errors

Most first-hand investigations involve taking measurements. Errors are mistakes that occur
in measurements or calculations. There is an error in every measurement you make
because the measuring instrument itself is never exact. Errors can also occur because an
instrument is not working properly or because a person is not using it properly. Errors also
occur when a mistake is made reading a scale eg parallax error-caused by not looking
straight at the measuring scale.

2. Accuracy

Accuracy relates to how close your observed or measured value is to the correct/real value.
Accuracy is determined by the quality of the measuring device used. The more sensitive the
measuring device is to changes in the environment, the more accurately you can measure
the changes.
The scales used on any instrument are marked off into finer and finer divisions (lines). The
finest (smallest) division used limits the accuracy of the instrument
Accuracy can sometimes be improved by changing the experimental technique or the
equipment used to measure data.

3. Reliability

Reliability refers to whether results are repeatable by you or any other person who tries to do
your experiment. If an experiment can be done many times with similar results, the results
are said to be reliable. This consistency of results gives you confidence that you are not
making random errors when carrying out the procedure.

Taking repeat measurement also helps you identify anomalous results– those that are
clearly out of line with the other data. Such results might then be omitted or the results might
be averaged to reduce the impact of a single anomalous result.
4. Validity

A valid test refers to whether an experiment tests what it says it is testing. A valid test must
have all variables controlled (kept the same) except for the variable being tested (the
independent variable).

A control group is often necessary to compare with the treatment group in which the
independent variable is changed. It allows a scientist to be sure that any differences in the
results have been caused by the variable being tested.

An experiment cannot be valid if it is not designed properly to actually measure what it sets
out to measure.

Results of an experiment and the conclusion draw from these cannot be valid unless the
experiment has been design properly and the results gained are BOTH accurate and
reliable.

Questions about first-hand data to determine whether it is accurate, reliable and

valid

Accuracy  Is my equipment sensitive enough to detect the changes I am trying to

measure?
 Is there a digital device that would be more sensitive to the changes
to be measured ? eg. probe, digital thermometer etc.
 Would a measuring cylinder , pipette or burette be more appropriate
to measure a particular volume than a beaker
 Have measurements been made carefully and human error
minimized eg. have measures been taken to minimize parallax error?
 Has the equipment been used correctly
 Has data be recorded to enough decimal places?
 If any of these points are of concern then fixing these up could
increase the accuracy of results. This will in turn increase the validity
of the data.

Reliability  Have I tested with repetition? Do I need to test a larger sample size?
 Increasing sample size in itself doesn’t make the results reliable but
gives the opportunity to gain reliable results if it is apparent that the
results achieved by following the same method are consistent.

Validity  Does my procedure actually test the hypothesis? What variables have
I identified and controlled?
 Is a control necessary? Does my conclusion relate to my results?
 Are there any factors that are not or cannot be controlled and which
may
 Have impacted the results. (so something that brings into question
that it was the IV the affected the result noted in the DV)
 Are my results both reliable and accurate

Example 1
 Kim wanted to find out whether salt (sodium chloride) raises the boiling point of water: Pure
water boils at 100°C at normal atmospheric pressure and temperature

Kim carried out the following procedure: She dissolved 5 g of salt in 100mL of distilled water,
heated the solution and recorded the boiling point with a thermometer that had divisions of
2°C. It was 100°C. Kim concluded that salt does not elevate the boiling point of water. Was
Kim right?

Apart from the fact that Kim only carried the experiment out once (so could not check if her
result was reliable) she also assumed that there would be enough of a rise in temperature
above 100°C for the thermometer to detect it. However, the thermometer chosen may not
have been sensitive enough to show the temperature change.

The more sensitive the measuring device is to changes in the environment, the more
accurately you can measure the changes.

Kim repeated the experiment using a temperature probe and data logger. The probe can
detect changes as small as 0.2°C. The boiling point of the salt solution was 100.8°C. She
repeated it three times and obtained the same result each time.

Kim concluded that salt does raise the boiling point of water. She had confidence in her
conclusion because, by repetition, she established a consistent pattern of results for the
same experiment. Several other students did the experiment with different probes and data
loggers (same sensitivity as the one used above) and confirmed the pattern - a 0.8°C rise.

The first hand investigation was repeated with three mercury thermometers with divisions of
0.2°C and two alcohol thermometers with divisions of 0.1°C. Similar results were found

The consistency of the results, regardless of how they were measured, gave Kim confidence
that her results were reliable. Kim concluded that salt elevates the boiling point of water.

However, was Kim’s procedure a valid test? NO!

It was not a valid procedure because we cannot be certain that the salt raised the boiling
point to 100.8°C. Perhaps there were impurities in the water which elevated the boiling point
to 100.8°C. The distilled water may not have been made properly. Perhaps salt had nothing
to do with elevating the boiling point of the water!

To rule out this possibility, the procedure must include an experimental control. This would
involve measuring the boiling point of the water source without the salt.

We also assume that Kim controlled all other variables.

Kim measured the boiling point of the distilled water without the salt. She used the
temperature probe as well as thermometers with divisions of 0.1°C and 0.2°C. Kim found
that the boiling point of the distilled water was 100°C.

Kim correctly concluded that salt does elevate the boiling point of water!

Questions
1. What did Kim do to make sure that her results were as accurate as possible?

2. What steps in the procedure helped Kim to be confident of the reliability of her results?

3. Kim’s first hand investigation had accurate and reliable results, however they were not valid.
Explain.

4. Can the results of an experiment be valid if they are not reliable? Explain.

Questions about first-hand data to determine whether it is accurate, reliable and valid

Accuracy Is my equipment sensitive enough to detect the changes I am trying to measure?

Is there a digital device that would be more sensitive to the change being measured?
eg. probe, digital thermometer etc.

Would a measuring cylinder , pipette or burette be more appropriate to measure a

particular volume than a beaker ?

Have measurements been made carefully and human error minimized eg. have
measures been taken to minimize parallax error?

Has the equipment been used correctly

If the answer to any of these is NO then fixing these up could increase the accuracy of
results. This will in turn increase the validity of the data.

Reliability Have I tested with repetition? Do I need to test a larger sample size?

Increasing sample size in itself doesn’t make the results reliable but gives the
opportunity to gain reliable results if it is apparent that the results achieved by following
the same method are consistent.

Validity Does my procedure actually test the hypothesis? What variables have I identified and
controlled?

Is a control necessary? Does my conclusion relate to my results?

Are there any factors that are not or cannot be controlled and which may have
impacted the results. (bringing into question if it was indeed the IV that affected the
DV)

Are my results both reliable and accurate

Example 2
Charlotte w anted to compare the amount of air in two brands of ice-cream. She placed a
large spoon of each ice-cream in two different cups and let the ice-cream melt. She then
measured how much liquid was in each cup. There was less liquid in cup B so she
concluded that ice-cream B must contain more air.

1. Outline how the accuracy of Charlotte’s experiment could be increased.

2. Outline how the experiment could be made more reliable.

3. Describe one way that Charlotte could improve the validity of her experiment.

Match each term in the left hand column with its definition in the right hand column by
placing its corresponding letter in the column to the left of the definition

A Accuracy Refers to whether the experiment is testing what it says it is

testing. Is the hypothesis being tested correctly?

B Reliability a mistake made whilst measuring or calculating

C Validity of the Data that is accurate, reliable and obtained from a valid
experiment experiment

D Valid data things that can change in an experiment

E Error this is identical to the experiment/treatment group except for

the one variable being tested(the IV)

F Variables refers to how close a measurement is to the actual result

using the best measuring instrument available

G Control group refers to whether results are repeatable (show a consistent

pattern)