Field Experiments in sociology

Last Updated on August 21, 2026 by Karl Thompson

Field experiments take place in real-life social settings such as a school, a workplace or the high street, rather than in a laboratory. This post covers:

1. What a field experiment is, and how it differs from a laboratory experiment
2. Seven examples of field experiments relevant to A-level sociology, from 1924 to 2025
3. The practical strengths and limitations of the method
4. The ethical strengths and limitations
5. The theoretical strengths and limitations
6. Rosenthal and Jacobson’s 1968 study in full — the one you are most likely to need
7. How field experiments come up in the AQA exams, including Methods in Context

You need to know about field experiments for the research methods component of A-level sociology, and the AQA does seem to like setting questions on experiments.

What is a field experiment in sociology?

An experiment measures the effect of an independent variable (the cause) on a dependent variable (the effect). The researcher changes one thing, holds everything else as steady as they can, and measures what happens.

A laboratory experiment does this in an artificial, controlled environment. A field experiment does it in a naturally occurring social setting — a classroom, a factory, a park, a recruitment process. The people being studied are usually going about their ordinary lives, and often do not know an experiment is happening at all.

Field experiments are far more common in sociology than laboratory experiments. Sociologists hardly ever use lab experiments, because the artificial environment of the laboratory is so far removed from real life that most sociologists think the results tell us very little about how people would actually behave outside it. If you want to know how teachers treat pupils, you have to go where the teachers and pupils are.

It is actually quite easy to set one up. Say you wanted to measure the effect of a teaching method on educational performance. You would get teachers to give a short test to establish current attainment levels, then change one aspect of their teaching for one class but not for a comparable class, run that for a term, and test everyone again at the end. Teaching method is your independent variable; test scores are your dependent variable; the class that carries on as normal is your control group.

Field experiments compared with laboratory experiments

	Laboratory experiment	Field experiment
Setting	Artificial, controlled	Real social setting
Control over variables	High	Low
External validity	Low	High
Awareness of being studied	Participants always know	Participants often do not know
Typical scale	Small	Can be very large
Favoured by	Positivists	Sociologists of most persuasions, when experiments are used at all

The trade-off is the thing to remember, and it is the point most 10-mark answers turn on: you buy realism by giving up control. A laboratory lets you isolate one variable precisely, but the setting is nothing like real life. A field experiment happens in real life, but you can never be sure that the thing you changed is the thing that produced the result.

One warning while we are here. The Stanford Prison Experiment is not a field experiment, however often it turns up on lists of them. A mock prison built by researchers in a university basement is an artificial setting, which makes Zimbardo’s study a laboratory experiment — a simulation, to be exact. Offer it as a field experiment in an exam and you lose the mark.

Seven examples of field experiments in sociology

Field experiments are not the most widely used method in sociology, but examiners love asking about them, and a specific example is worth far more in an exam than a general definition. Below are seven, arranged as a timeline with the most recent first.

The first two are large, recent and British, which makes them the most useful ones to have at your fingertips. The older ones are the classics you will meet in every textbook.

2025 — The BITUP school attendance trial

School absence is the biggest single problem in English education at the moment, and this is the best recent example of sociologists and behavioural scientists using a field experiment to test a solution.

The Behavioural Insights Team designed a simple intervention. At the start of each half-term, schools sent a personalised text message to the parents of any pupil whose attendance had dropped below 95%, telling them the number of days their child had missed rather than a percentage — “Tom missed five days of school last term” instead of “Tom’s attendance was 90%”. The thinking was that parents underestimate absence when it is expressed as a percentage, and that the start of a half-term offers a “fresh start” at which new habits can form.

[IMAGE: needed — suggest a simple graphic contrasting “attendance 90%” with “missed 5 days of school last term” · alt text: “BITUP text message field experiment on school attendance for A-level sociology”]

The design is worth knowing in itself. Randomisation happened at family level within schools, so intervention and control families sat side by side in the same school. That removes the risk that differences between schools explain the result. At randomisation the trial covered 87,909 families and 104,029 pupils across secondary schools in England; 105 schools were in the final analysis, and the intervention ran through the 2023–24 school year.

The finding: pupils in the intervention group were absent for 0.21 fewer days across the whole academic year than pupils in the control group — an increase of 0.10 percentage points in the average attendance rate. The effect was larger for pupils eligible for free school meals (0.51 fewer days), for girls (0.35) and for Year 8 (0.61), though those subgroup findings are less secure because the numbers are smaller. The intervention cost £1.11 per pupil per year.

HOWEVER, read what the EEF itself said about it: the reduction is small compared with what would be needed to shift attendance meaningfully, so on its own the approach is not highly promising. The evaluation also found no conclusive evidence that the texts changed what parents knew about their child’s absence or what they thought about attendance — which undermines the theory the intervention was built on, even though the behaviour shifted slightly.

Why it matters: this is the single best example on the list of what a properly designed field experiment can and cannot establish. The randomisation is clean, the sample is enormous, the finding carries a high security rating — and the effect is tiny. A student who can say “the strength of the method is that we can trust this small result, where a survey could not have told us whether the texts caused anything at all” is doing real evaluation.

Source: BITUP: Updating Parents on Number of School Days Missed, Education Endowment Foundation and Youth Endowment Fund, evaluation by Verian, report published October 2025.

2022 — The UK four-day week pilot

Between June and December 2022, 61 UK organisations and around 2,900 workers moved to a four-day week on full pay. The model was 100:80:100 — 100% of pay, 80% of the hours, on the expectation of 100% of the output. The trial was coordinated by 4 Day Week Global with the Autonomy Institute and the 4 Day Week Campaign, with research led by Juliet Schor at Boston College and Brendan Burchell and David Frayne at Cambridge. Participating organisations ranged from a robotics manufacturer to a fish-and-chip shop in Norfolk.

[IMAGE: needed — suggest a simple chart of the headline outcomes: 92% continued, 71% reduced burnout, 57% fall in staff leaving · alt text: “UK four-day week pilot results field experiment for A-level sociology”]

The results, published in February 2023, were emphatic. Of the 61 companies, 56 (92%) carried on with the four-day week and 18 made it permanent. 71% of employees reported reduced burnout and 39% reported less stress. The number of staff leaving fell by 57% and sick days by 65%, while revenue held broadly steady.

A follow-up in February 2024 found at least 54 of the 61 organisations still operating the policy, with 31 having made it permanent. A further pilot of 17 British businesses in late 2025 reported that every one of them kept the four-day week, with 62% of staff reporting reduced burnout.

Why it matters: this is directly usable well beyond methods. It bears on work and leisure, on the domestic division of labour — a fifth day at home does not automatically mean men do more of it, which is exactly the sort of question feminist sociologists ask — and on the argument that longer hours do not produce more output.

HOWEVER, and this is the point that earns marks: it is not a controlled experiment. The companies volunteered, so the sample selects for organisations already sympathetic to the idea and confident they could manage it. There was no control group of comparable firms carrying on as normal, so there is no way to separate the effect of the shorter week from everything else happening in those workplaces over six months. And the research was commissioned by organisations campaigning for a four-day week, which raises a fair question about value freedom.

Set that against BITUP and you have the whole methodological argument in one pair of examples: the study with proper randomisation found a tiny effect it can defend; the study without it found a huge effect it cannot.

Source: Lewis, K., Stronge, W., Kellam, J. et al. (2023) The Results Are In: The UK’s Four-Day Week Pilot, Autonomy Institute. Follow-up: Autonomy, February 2024.

2014 — The domestic abuse in the lift experiment

A Swedish group set up a hidden camera in a lift while actors played an abusive boyfriend and his victim. The male actor swore at and physically assaulted the woman while ordinary members of the public shared the lift with them.

Only one person out of 53 intervened.

The experiment was staged by STHLM Panda, a YouTube channel which describes itself as documenting the society we live in. It was widely reported, including by the Guardian.

Why it matters: it is a real-world illustration of the bystander effect, and it is useful for the domestic violence section of families and households as well as for methods. HOWEVER, note what it is — a piece of video content made by a YouTube channel, not a study with a published methodology. You cannot check the sampling, you do not know how much footage was discarded, and 53 people is a small number. Use it as an illustration, and say so; do not present it as though it were peer-reviewed research.

2010 — The ethnicity and bike theft experiment

Three actors, dressed similarly and equipped with bolt cutters, took turns openly trying to saw the lock off a bicycle chained to a post in a public park. The only variable that changed was who was doing it.

[VIDEO EMBED: retain existing — bike theft experiment]

When a white male actor did it, roughly 100 people walked past over the course of an hour and only one couple tried to stop him. Several passers-by asked whether it was his bike, accepted “not exactly” as an answer, and carried on.

When a black male actor did exactly the same thing, a crowd gathered within minutes, people photographed him on their phones as evidence, and the police were called.

When an attractive young blonde woman did it, several men stopped to help her.

Why it matters: this is an unusually clean demonstration of racial and gendered stereotyping in the perception of crime, and it connects directly to the labelling theory of crime and to debates about stop and search. It also shows how a single-variable design can make a point that survey data struggles to.

Provenance, and this matters: the footage comes from ABC’s What Would You Do?, an American hidden-camera television programme presented by John Quiñones, first broadcast in May 2010. It is a television format, not a research study, and it was filmed in the United States rather than the UK. Say that in an essay and you have turned a weak example into an evaluation point.

2009 — The ethnicity and job application experiment

Researchers at the National Centre for Social Research, commissioned by the Department for Work and Pensions, sent 2,961 fictitious job applications to 987 real advertised vacancies between November 2008 and May 2009. Three applications went to each vacancy, closely matched for education, skills and work history, and differing only in the name at the top — names recognisably associated with different ethnic groups, including Nazia Mahmood, Mariam Namagembe and Alison Taylor. Every fictitious applicant had a British education and a British work history, so nothing else could plausibly explain a difference in treatment.

Nine occupations were covered, from accountants and IT technicians to care workers and sales assistants, across Birmingham, Bradford, Bristol, Glasgow, Leeds, London and Manchester.

[IMAGE: retain existing — cv.png · caption: “Boleslav would be twice as likely to get an interview if he called himself Brian”]

An applicant with a white British name had to send nine applications to receive one positive response. An equally qualified applicant with an ethnic minority name had to send sixteen. In net terms, the researchers found 29% discrimination in favour of white names.

The report concluded that there was no plausible explanation for the difference other than racial discrimination.

One further finding is worth knowing, because it is more precise than the version usually repeated: applications made through employers’ own standard application forms showed virtually no discrimination at this stage, while applications made by CV did. Since public sector vacancies were the ones most likely to use standard forms, public sector employers appear less discriminatory in the raw figures — but the mechanism is the form, not the sector.

Source: Wood, M., Hales, J., Purdon, S., Sejersen, T. and Hayllar, O. (2009) A Test for Racial Discrimination in Recruitment Practice in British Cities, DWP Research Report 607. Also covered by the Guardian and checked by Full Fact.

Why it matters: this is the example to use when you need a field experiment with a large sample, a published methodology and UK data. It is directly relevant to ethnicity and inequality, to the labour market, and to the argument that discrimination is structural rather than a matter of individual prejudice.

1968 — Rosenthal and Jacobson’s self-fulfilling prophecy experiment

Robert Rosenthal and Lenore Jacobson set out to measure the effect of high teacher expectations on pupils’ educational performance. This is the field experiment you are most likely to need, and it gets a full treatment further down this post.

In short: they gave pupils at a California primary school an IQ test, then told teachers that 20% of them had been identified as likely to “spurt” academically over the coming year. In reality that 20% had been picked at random. Eight months later, all the pupils were retested. The average pupil had gained 8 IQ points; the “spurters” had gained 12.

[IMAGE: retain existing — self-fulfilling-prophecy.jpg · alt text: “Rosenthal and Jacobson self-fulfilling prophecy field experiment for A-level sociology”]

Why it matters: it is the empirical foundation of labelling theory and the self-fulfilling prophecy in the sociology of education, and it is the standard example for Methods in Context questions about researching teacher expectations.

1924–32 — The Hawthorne factory experiments

Western Electric’s Hawthorne Works in Cicero, Illinois, on the edge of Chicago, was the site of a long series of studies into what made workers more productive. Researchers varied lighting levels, rest breaks, the length of the working day, refreshments and the layout of workstations, and measured output.

[VIDEO EMBED: retain existing — Hawthorne experiments]

Productivity went up after almost every change. Then it went up when changes were reversed. At one point the special conditions were withdrawn entirely and the women in the test room returned to a full 48-hour week with no rest breaks — and output reached its highest level yet recorded. When the study ended, productivity slumped.

The conclusion drawn was that the workers were responding not to the lighting or the breaks but to the attention: they knew they were being studied, and they behaved differently because of it. This gave sociology the term the Hawthorne effect, meaning any short-term change in behaviour caused by participants knowing they are taking part in research rather than by the independent variable.

Two details worth getting right, because most revision material gets them wrong. The famous illumination experiments ran from 1924 to 1927 and were run by the company’s own engineers with the National Research Council. Elton Mayo and the Harvard team, whose names are attached to the studies in most textbooks, arrived in 1927–28, after the puzzling results had already appeared. And the whole programme was much larger than a single field experiment — the later phases involved interviewing some 21,000 employees.

Why it matters: the Hawthorne effect is one of the most useful concepts in the whole methods topic, because it applies to almost every method where respondents know they are being researched. It is also the standard criticism of any experiment. HOWEVER, it is worth knowing that later re-analyses of the original illumination data have questioned whether there was a measurable Hawthorne effect at Hawthorne at all — which makes the concept a nice example of how a finding can outlive the evidence for it.

Is this sociology or psychology?

A reader once left a comment on the older version of this list saying that most of these were psychology rather than sociology. It was a fair challenge at the time and worth answering, because the distinction comes up in exams.

Sociologists rarely run experiments, so there are not many purely sociological ones to point at, and we have always borrowed from our sister subject. What makes a study sociological is not who ran it but what it is about. BITUP is about how schools and families interact around attendance; the four-day week pilot is about the organisation of work; the DWP recruitment study is about structural discrimination in a labour market; Rosenthal and Jacobson is about how institutions sort children. Those are sociological questions, whoever holds the clipboard.

The examples that sit least comfortably here are the two pieces of television. They are about individual reactions in the street rather than about institutions, and they were made to be watched rather than analysed. Use them, but know what they are.

Practical strengths and limitations

The rest of this post uses the PET framework — practical, ethical and theoretical — which is how the AQA expects you to evaluate any research method, and how methods questions are usually structured.

Field experiments can be run at a scale laboratories cannot manage. You cannot get a school or a factory into a laboratory. You can run an experiment inside one, involving thousands of people who are there anyway. BITUP covered more than 100,000 pupils across a hundred-odd schools; the DWP recruitment study covered 987 employers in seven cities for little more than the cost of postage and researcher time.

They can piggyback on things institutions already do. Schools trial new teaching approaches, streaming arrangements and interventions all the time, and if two comparable groups are treated differently and both are assessed, that is a field experiment whether anyone calls it one or not. This makes existing school data a rich source for researchers — see my post on experiments within schools.

Some designs are extremely cheap. BITUP cost £1.11 per pupil per year, which is one reason a policy trial at that scale was possible at all.

HOWEVER, access is the standing practical problem. Schools and workplaces are far more reluctant to let a researcher in than a university is to lend out a room. Gatekeepers — headteachers, managers, governing bodies — can refuse, or agree and then restrict what you may do. Three schools withdrew from BITUP after recruitment. A laboratory has no gatekeeper.

Delivery depends on the institution, not the researcher. BITUP’s evaluators found that schools did not always follow the guidance about which families to exclude, and that some schools’ messaging software made the intervention much harder to run than others. In a laboratory the researcher administers the treatment; in the field, someone else does, and they have a day job.

They take a long time. Rosenthal and Jacobson waited eight months for their second round of testing. BITUP ran across a full academic year. A field experiment often cannot be hurried, because the process being studied takes as long as it takes.

The setting does not stay still. Staff leave, pupils move schools, a workplace reorganises, the economy changes. The DWP study ran during a recession, which the researchers themselves noted may have affected employers’ behaviour.

Ethical strengths and limitations

The main ethical strength is that participants are usually in a setting they chose to be in. Nobody is confined, isolated, or put through an ordeal constructed for the purpose.

Some designs cause almost no harm to individuals at all. In correspondence tests like the DWP study, the people being tested are employers acting in a professional capacity, and no real applicant is disadvantaged. That is about as ethically clean as covert research gets.

And some are genuinely benign. BITUP tested whether a text message helped. The worst that could happen to a family in the control group was that they carried on being contacted by the school in the usual way.

HOWEVER, deception and lack of informed consent are close to unavoidable in most designs. This is the central ethical problem with the method, and it is structural rather than incidental. If participants know what is being tested, the Hawthorne effect wrecks the design — which is precisely what the Hawthorne studies themselves demonstrated. So researchers deceive: Rosenthal and Jacobson lied to teachers about a test that did not exist; the bystanders in the lift believed they were watching a real assault.

Withholding a treatment raises its own problem. BITUP was open about what it was doing, but half the families were deliberately not sent messages that the researchers hoped would help their children attend school. That is the standard ethical objection to any randomised trial, and it is a fair one — though the answer is also fair: nobody knew whether the messages worked, which is exactly why the trial was run.

Harm is a genuine risk in the older studies. The “spurters” seem to have benefited from Rosenthal and Jacobson’s study; the other 80% of pupils did not, and may have been held back if teachers redistributed attention towards the group they had been told was gifted.

Debriefing is often impossible. You can debrief a laboratory participant. You cannot debrief 53 anonymous lift passengers or the hundreds of employers who binned a fictitious CV.

Most of the classics could not be run today. Ethical review is far stricter than it was, and child welfare in particular is taken far more seriously than in 1968. The British Sociological Association’s statement of ethical practice sets out current expectations. That is itself a useful point: the reason the recent examples on this list look so much tamer than the old ones is partly ethical.

Theoretical strengths and limitations

External validity is the headline strength. Because the setting is real, the behaviour is real. Whatever a teacher does when a researcher is not watching, they were doing it in Oak School. Whatever an employer does with a CV, they did it with these CVs. This is what laboratory experiments cannot deliver, and it is why sociologists prefer field experiments when they use experiments at all.

A well-designed field experiment gives unusually strong evidence of cause. Survey data can show that ethnic minority applicants get fewer interviews; it cannot rule out differences in qualifications, experience or the jobs applied for. A correspondence test can, because everything except the name is held constant. Keizer, Lindenberg and Steg’s six field experiments in Groningen make the same point from a different direction: by manipulating litter and graffiti and then measuring whether people stole an envelope, they produced causal evidence for broken windows theory that no amount of crime survey data could have given.

Large-scale designs can be genuinely representative. BITUP covered over 100,000 pupils; the DWP study nearly 3,000 applications across nine occupations and seven cities.

HOWEVER, extraneous variables cannot be controlled. This is the standing theoretical weakness. Rosenthal and Jacobson claimed higher teacher expectations produced the higher scores, but they never observed classroom interaction, so they had no evidence for the mechanism they were proposing. Something else in those eight months may have produced the difference.

Without random assignment and a control group, it is not really an experiment at all. The four-day week pilot is the clearest illustration. Volunteering companies, no comparison group, a six-month window in which anything might have changed — the striking findings could be the shorter week, or self-selection, or the Hawthorne effect operating on an entire workforce that knew it was part of a famous trial. BITUP, which had proper randomisation, found an effect one-hundredth as dramatic. Which of those two is closer to the truth about what interventions actually achieve is a question worth asking in an essay.

Reliability is limited. The design may be simple to describe, but the exact conditions cannot be reproduced — schools differ, intakes differ, cohorts differ. Repeating a field experiment gives you a similar study, not the same one.

Positivists prefer laboratories, and Interpretivists dislike experiments altogether. Positivists want control, precise measurement and replicability, and a field experiment gives up all three to a degree. Interpretivists object more fundamentally: experiments measure behaviour without accessing meaning, and treat people as objects reacting to stimuli rather than as conscious actors who interpret their situations. BITUP is a neat case in point — it measured a behavioural change but found no evidence that parents’ understanding had shifted at all, which is precisely the gap an Interpretivist would say experiments cannot see into.

Neither of the two main theoretical traditions is fully comfortable with the method, which is a large part of why it is so rarely used.

Rosenthal and Jacobson’s field experiment in full

Because this is the study you are most likely to be asked to write about, here it is properly.

Aim

To measure the effect of high teacher expectations on the educational performance of pupils.

Procedure

Rosenthal and Jacobson carried out their research in 1965 at a California primary school they called “Oak School”. All the pupils sat an IQ test, which the researchers presented to teachers as a new instrument — the “Harvard Test of Inflected Acquisition” — capable of identifying children about to make unusual intellectual progress. On the basis of it, teachers were told that 20% of pupils were likely to “spurt” academically over the coming year.

In reality the 20% had been selected at random, and no such test existed. There was no reason to expect those children to progress faster than anyone else — unless the teachers’ beliefs about them made a difference.

All pupils were retested eight months later, and again after a further year.

Findings

Over the first eight months, pupils gained an average of 8 IQ points. The “spurters” gained 12.

The gains were concentrated among the youngest children, aged six to eight. After the further year, the expectancy advantage showed up mainly among the ten- and eleven-year-olds.

Rosenthal and Jacobson concluded that teachers’ expectations had been passed on to pupils through the way teachers interacted with them, producing a self-fulfilling prophecy. The findings were published in 1968 as Pygmalion in the Classroom.

Evaluating the study

Deception and lack of informed consent. For the design to work, teachers had to be misled about what was being tested, and the pupils had no idea anything was happening.

Harm to the control group. The spurters appear to have benefited. The remaining 80% did not, and may have been disadvantaged if teachers gave disproportionate attention to the group they believed was gifted. Given how central child welfare is to schools today, it is unlikely this study would be approved now.

No evidence for the mechanism. Rosenthal and Jacobson claimed teacher expectations were transmitted through classroom interaction, but they conducted no classroom observation. Claiborn (1969) and others subsequently found no evidence of expectations being passed on in the way they described. The causal story is an inference, not a finding.

Measurement problems. Robert Thorndike, an expert on psychological testing, attacked the study almost immediately, arguing that the IQ instrument was inappropriate for children this young and that some of the initial scores were implausibly low — which would inflate the apparent gains on retesting.

The effect has shrunk under replication. The original post noted that the study was repeated 242 times within five years; I have not been able to verify that specific figure, but it is certainly true that a very large number of replications followed. What those replications found is the important part: Raudenbush’s 1984 meta-analysis put the average teacher-expectancy effect at around d = 0.14, and Jussim and Harber’s 2005 review concluded that classroom expectancy effects are real but small, and much smaller than Pygmalion implied.

Reliability. The design is simple enough to repeat, but the exact conditions are not reproducible, given the differences between schools and intakes.

That combination — a famous finding, weak on its own terms, that turned out to point at something real but much smaller — is exactly what examiners mean by evaluation. A student who can say that is doing better than one who simply reports 12 points against 8. It is also, incidentally, the same shape as the BITUP result: an intervention that does something, but far less than its advocates hoped.

Field experiments in the AQA exams

Field experiments come up in two distinct places, and they are assessed differently.

Research methods on Papers 1 and 3

Experiments are examinable as part of Theory and Methods, which appears on Paper 1 and, more substantially, on Paper 3. Questions here are about the method itself: its strengths, its limitations, and the theoretical debates around it.

The reliable structure for these answers is PET. Take practical, ethical and theoretical in turn, make a point in each, and support each one with a named example rather than a generic statement. “Field experiments raise ethical problems” earns very little. “Rosenthal and Jacobson had to deceive teachers about a test that did not exist, and the 80% of pupils not labelled as spurters may have received less teacher attention as a result” earns considerably more.

If you can pair a recent example with a classic one, do. Examiners notice.

Methods in Context on Paper 1

This is the 20-mark question on Paper 1 that asks you to apply a method to a specific issue in education. Field experiments come up regularly, and you now have two anchors: Rosenthal and Jacobson for teacher expectations, and BITUP for school attendance.

The mark scheme rewards application to the educational context specifically, not general methods knowledge. So do not write about access in the abstract — write about headteachers as gatekeepers, about parental consent for research involving children, about the fact that classes are not randomly assigned so your groups may differ before you start, about pupils moving schools mid-study, and about the school staff who have to deliver the intervention on top of everything else they do. BITUP gives you all of these with evidence attached.

See my post on research methods in context: experiments and education for a fuller treatment.

Theory and Methods A Level Sociology Revision Bundle 

If you like this sort of thing, then you might like my Theory and Methods Revision Bundle – specifically designed to get students through the theory and methods sections of  A level sociology papers 1 and 3.

Contents include:

• 74 pages of revision notes
• 15 mind maps on various topics within theory and methods
• Five theory and methods essays
• ‘How to write methods in context essays’.

Signposting and Related Posts

This post has been written primarily for students studying research methods as part of the AQA’s A-level sociology course.

Before this topic, most students cover An Introduction to Experiments in Sociology, which explains hypotheses, independent and dependent variables and the basic logic of experimental design.

The natural companion post is Laboratory Experiments in Sociology, which covers the other main type and why sociologists rarely use it.

If you came here looking for The Stanford Prison Experiment, that post explains why it is a laboratory experiment rather than a field one, and why its findings are no longer accepted as they once were.

For a much longer list of examples, including some unusual contemporary ones, see Sociological Experiments.

Are Chinese Teaching Methods the Best? is a field experiment in tough teaching methods conducted in a UK school in 2015 — another useful education example.

Experiments Within Schools covers the small-scale experiments schools run on themselves, which are a rich source of material for Methods in Context.

For condensed revision notes covering both types, see Experiments in Sociology — Revision Notes.

For everything else on this topic, see the main research methods page.

Please click here to return to the homepage – ReviseSociology.com