Last Updated on October 13, 2026 by Karl Thompson
Experiments are the only research method that lets educational researchers measure the effect of one specific variable on pupils’ behaviour or results. In education, they come in four types:
- Field experiments run by schools themselves, testing things like setting, class composition or new teaching techniques on one class before rolling them out.
- Field experiments run by researchers inside schools, the classic example being Rosenthal and Jacobson’s Pygmalion in the Classroom.
- Natural experiments, where a policy change (such as academies and free schools) or an event (such as Covid school closures) creates groups that can be compared.
- Laboratory experiments, which are rare in the sociology of education because the setting is so artificial.
This post gives examples of all four, followed by the practical, ethical and theoretical strengths and limitations of using experiments to research education. It has been written primarily for students studying the Methods in Context part of the AQA’s A-level sociology course, where you have to apply a research method to an education topic. For how to answer the 20-mark Methods in Context question itself, see How to Answer Methods in Context Questions.
Field experiments run by schools themselves
Experiments conducted by schools themselves are far more common than educational researchers conducting their own field experiments in schools. If you have been through thirteen years of formal education, you have probably been part of one at some point, whether you knew it or not.
It is quite usual for schools to run small-scale field experiments to try out a new teaching technique, or to see whether banding or streaming works. If there are several classes of similar ability doing the same subject, it is relatively easy to keep some classes being taught in the usual way, change one aspect of the teaching for another class, and measure what difference it makes to behaviour or learning.
An example from my own college
Some years ago we ran the following experiment in my institution, designed to test whether splitting students into ability bands improved results.
In five subjects we deliberately created one class made up entirely of higher-ability students (those predicted Bs or above at A-level on the basis of their GCSE grades), while keeping some high-ability students in each of the regular mixed-ability classes. In A-level sociology, for example, we had seven classes at AS, so we ended up with one “high ability” class and six mixed-ability classes, each with two or three A/B students rather than the three to five they would otherwise have had.
The hypothesis of the senior management was that grouping the higher-ability students together would improve their results. The control group was the high-ability students who stayed in the regular mixed classes.
We let this run for a year and compared the AS results at the end (it was a long time ago, when we still had AS exams). The results were inconclusive: across all five subjects, there was no significant evidence that putting the top-band students in one class led to better results.
As far as I am aware, the students involved still have no idea they were part of an experiment.
This kind of experiment is more common than you think
A lot of schools put students into ability bands, and it makes sense that they review the results from time to time and experiment to see whether mixing the bands up gives better outcomes. If you have six maths groups in a year, for example, you could have six discrete bands, or two broader bands with a wider range of ability spread across three classes each.
More generally, schools can change one variable in one class but not in others, just to see whether it makes a difference over a year, rather than risk rolling a change out across a whole year group. Variables a school might change include:
- the gender mix of classes
- seating layouts within classrooms
- the length of lessons, or the number of lessons a week
- the timing of support lessons: before or after school, at weekends, in the holidays
- how support staff are used in classrooms
- wider school policies on uniform, discipline and punishment
- the teaching techniques used in lessons
- the balance between face-to-face and online learning
Strengths of schools conducting their own experiments
Don’t kid yourselves: this kind of micro-experiment goes on all the time, and there are some good reasons for it.
Ethically, teachers and schools have a duty to give students the best education they can. Educational theory about what works changes, and technology changes, so schools have to adapt. Trying something with one class for a year is a sensible way of finding out how to introduce a change across the whole school the following year, and if the change doesn’t work, at least it isn’t every student who suffers.
Practically, the students are there and the school is there, so it is relatively easy for a school to run an experiment, compared with an outside researcher having to negotiate access.
Theoretically, validity should be high, because students are typically not told they are part of an experiment, so there is little reason for them to change their behaviour. Reliability should also be reasonable, because conditions in most schools are fairly stable over time.
Limitations of schools conducting their own experiments
Ethically, there is the problem of deception, and of some students being treated differently from others for the length of the experiment, which goes against equal opportunities.
Practically, teaching one class differently from the rest can be stressful and demanding for the teacher involved.
Theoretically, in very small schools, or subjects with only a few classes, it is hard to get a sample large enough to be representative.
Case study: the “Chinese School” at Bohunt
One of the most extreme recent field experiments in an English school was filmed for the BBC Two documentary series Are Our Kids Tough Enough? Chinese School, broadcast in 2015. Chinese teachers took over the education of a mixed-ability class of 50 Year 9 pupils at Bohunt School in Liphook, Hampshire, for four weeks, teaching them the way they would teach in China. At the end, the class was tested in maths, science and Mandarin, and their results were compared with pupils in the same year who had carried on with Bohunt’s usual teaching.
The main features of the “Chinese School” were:
- a 12-hour school day, from 7am to 7pm
- lessons that were mostly lectures, with the teacher at the front writing the theory on the board while students (were supposed to) take notes and learn
- compulsory PE, in which students were timed, tested and ranked against each other
Results. The Chinese-taught class did better, by around 10 percentage points on average. Local press reports of the final results put the class at nearly 68% in maths against nearly 55% for the comparison group, 58.5% against 50% in science, and just over 46% against just under 36.5% in Mandarin (Liphook Herald, 2015).
The experiment also revealed something of a culture clash. Students who were not particularly self-disciplined or well-behaved did not respond well to a style of teaching that is less student-centred and less inclined to encourage individualism.
Limitations of the Chinese School experiment
- Control and comparison. We aren’t told much about how the two groups compared at the start, or exactly how much each improved. I’m assuming this was controlled for, but a documentary isn’t a research report.
- The Hawthorne effect. The improved results might reflect the students knowing they were part of an experiment (and knowing they were on TV), or simply being exposed to something different, rather than anything specific to Chinese teaching methods. A change is as good as a rest!
- Representativeness. Bohunt seems to be an unusually good, outward-looking school, which is reflected in it agreeing to take part at all. Would you get the same findings somewhere else?
- Ethics. Some (wrong) individuals might argue that some of the children’s self-esteem was harmed by being ranked in PE; other (right) individuals might argue that this is just life, so get over it, kiddo.
Field experiments by sociologists within schools
Examples of sociologists going into schools to run their own experiments are much rarer. The classic is Rosenthal and Jacobson’s Pygmalion in the Classroom (1968), which set out to measure the effect of high teacher expectations on pupils’ performance.
At a Californian primary school they called “Oak School”, all the pupils sat an IQ test, which teachers were told could identify children about to make unusual intellectual progress. The teachers were then given the names of 20% of pupils who were likely to “spurt” in the coming year, when in reality those pupils had been chosen at random. When the pupils were retested eight months later, the “spurters” had gained more than the rest, especially among the youngest children.
Rosenthal and Jacobson concluded that this supports the self-fulfilling prophecy: teachers’ expectations were passed on to pupils through the way teachers treated them. HOWEVER, later attempts to repeat the experiment have produced mixed results, and the study raises serious ethical questions about deception and about the 80% of pupils who weren’t labelled as spurters.
For the full study and its evaluation, see Field Experiments in Sociology. For how it fits into the wider labelling debate in education, see Labelling Theory and the Self-Fulfilling Prophecy.
Natural experiments and education
A natural experiment is where the researcher doesn’t manipulate anything: a policy change or an event creates different groups, and the researcher compares them afterwards.
Two government policies have introduced new types of school in recent years: academies and free schools. We now have many years of data with which to compare their performance with schools that are still maintained by local authorities, which is effectively a natural experiment.
We could do the same at a global level, by looking at the PISA league tables and asking what the education systems of the top-performing countries have in common, if anything.
The Covid-19 school closures also created a natural experiment in remote learning. Research carried out by FFT Datalab for the Education Endowment Foundation, using test results from around 19,000 primary pupils, found that the gap between disadvantaged pupils and their classmates grew by about a month over the pandemic, mostly in maths during the first closures (EEF, 2022).
Be careful discussing natural experiments in the exam: in practice you are getting into secondary data here rather than pure experiments, although there are clear links between the two.
Laboratory experiments relevant to education
Laboratory experiments are rare in the sociology of education. The controlled, artificial setting of a lab is so far removed from a real classroom that most sociologists doubt the results tell us much about how teachers and pupils actually behave. For the general strengths and limitations of the method, see Laboratory Experiments in Sociology.
A more recent example, shown on the BBC Two series The Twinstitute, tested whether simply having a mobile phone on the desk affects test performance. Twins were split into two groups and sat the same IQ test under identical conditions, except that one group had their phones taken away and the other was told to leave their phones on the desk. The group with phones on the desk scored noticeably lower. For more on this, see Sociological Experiments.
Using experiments in Methods in Context questions
The 20-mark Methods in Context question appears on Paper 1, Education with Theory and Methods. It asks you to evaluate the strengths and limitations of one research method for investigating a particular education topic, applying material from the item. Any of the main methods can come up, experiments included, and the examples above give you the research contexts you need to apply the method properly rather than writing a general methods essay.
When you write about experiments in that question, think about who is being experimented on (pupils, teachers, whole classes), where (classrooms, whole schools, the home), and how sensitive the topic is. Then apply the practical, ethical and theoretical points from this post to that particular topic: the problems of experimenting on children, of getting consent from schools and parents, and of controlling variables in a real school.
For more on the research contexts themselves, see Researching Pupils in Education, Researching Teachers in Education and Researching in Classrooms.
Signposting and related posts
This post has been written primarily for students studying the Methods in Context part of the research methods topic, as part of the AQA’s A-level sociology course.
- Experiments in Sociology: An Introduction covers the key terms and the different types of experiment before you apply them to education.
- Field Experiments in Sociology gives the general strengths and limitations of the method, and Rosenthal and Jacobson in full.
- How to Answer Methods in Context Questions shows how to apply any method to an education topic in the exam.
- Research Methods for A-level Sociology is the main research methods page.
Please click here to return to the homepage – ReviseSociology.com