Positivism

Positivists hold that society can be studied with the logic of the natural sciences. Society exists outside the individual as a set of social facts that exert real force, and those facts can be measured.

• Patterns and trends. The positivist looks for correlations and causal laws in large datasets.
• Objectivity and value freedom. The researcher's own values must be kept out. Facts, not preferences.
• Quantitative data — numbers that can be compared, replicated and tested.

The classic study: Émile Durkheim's Suicide (1897) used official statistics from across Europe to show that suicide rates were stable and patterned — higher among Protestants than Catholics, among the unmarried than the married. Suicide, the most private of acts, turned out to have a social cause: variations in social integration and moral regulation. That, for positivists, is sociology proving itself a science.

Evaluation: Durkheim's data were only as good as the coroners who produced them. Interpretivists such as J. Maxwell Atkinson (Discovering Suicide, 1978) and Jack Douglas (The Social Meanings of Suicide, 1967) argued that suicide statistics are not facts at all but the outcome of officials' interpretations — coroners work with commonsense theories about what a 'typical suicide' looks like. If the statistic is socially constructed, Durkheim measured coroners, not suicide.

Interpretivism

Interpretivists reply that people are not billiard balls. Human beings attach meanings to their situations and act on those meanings — so explaining behaviour requires understanding it from the inside.

• Verstehen and empathy. Max Weber's term for the empathetic understanding of the meaning an actor gives to their own action.
• Rapport — trust built with participants, without which they will not tell you the truth.
• Subjectivity — the researcher's involvement is a resource, not only a contaminant.
• Researcher imposition — the danger that a pre-set questionnaire forces respondents into the researcher's categories, so the data reflect the sociologist's assumptions rather than the respondent's world.
• Reflexivity — the researcher openly examines how their own identity, presence and choices shaped the data.
• Qualitative data — rich, detailed, verbal.

Ann Oakley's essay 'Interviewing Women: a contradiction in terms' (1981) argued that the standard 'objective', detached interview is a masculine model that treats women as data-mines. She interviewed pregnant women repeatedly, answered their questions, formed friendships — and argued this produced more valid data, not less. This is the reflexivity point in its sharpest form.

Whose logic is it?

Validity, reliability, representativeness, generalisability

These four words are your evaluation engine. OCR expects you to use them evaluatively, not just define them.

• Validity — does the data give a true, authentic picture of what it claims to measure? Qualitative methods usually score high here: an unstructured interview lets people speak in their own terms.
• Reliability — could another researcher repeat the study using the same procedure and get the same results? Standardised methods (questionnaires, structured interviews) score high; a lone ethnographer's fieldwork is almost impossible to replicate.
• Representativeness — is the sample a typical cross-section of the population being studied?
• Generalisability — can the findings be extended beyond the sample? Only a representative sample supports generalisation.

The classic trade-off: methods that are high in reliability tend to be low in validity, and vice versa. A closed questionnaire is repeatable but shallow; a covert ethnography is deep but unrepeatable. The strongest answers show that no method wins on every criterion — which is exactly why triangulation exists.

Validity or reliability?

The research process

• Choosing the topic — not a free choice. It is shaped by the researcher's values, current theoretical fashions, and above all by funding: governments, charities and businesses pay for the questions they want answered.
• Aims, hypotheses, research questions. A positivist typically starts with a testable hypothesis; an interpretivist more often starts with an open aim, so as not to prejudge the participants' meanings.
• Operationalisation — turning an abstract concept into something measurable. How exactly will you measure 'poverty'? Peter Townsend (Poverty in the United Kingdom, 1979) operationalised it as relative deprivation, building an index of 60 indicators of an ordinary living standard. A different operationalisation gives a different poverty rate — which is why definitions are political.
• Pilot study — a small trial run to iron out confusing questions and check timings before committing resources.
• Primary data (you collected it) vs secondary data (someone else did).
• Respondent validation — taking your findings back to participants to check you have understood them. An interpretivist check on validity.
• Longitudinal studies — the same sample studied over years, so you can see change and cause rather than a single snapshot. Real examples: the Millennium Cohort Study, and J. W. B. Douglas's The Home and the School (1964), which followed over 5,000 children and helped establish the link between home circumstances and attainment. Weakness: sample attrition — people drop out, and the ones who stay may not be typical.
• Sociology and social policy — research can change the law and the welfare system, but governments commission and ignore research selectively.

Sampling, access and gatekeeping

You cannot study everyone, so you study a sample drawn from a sampling frame (a list of the population — an electoral register, a school roll, a membership list).

• Random techniques — random (equal chance for all), systematic (every nth name), stratified (divide the population into layers by class, sex, ethnicity or age, then sample each in proportion). These maximise representativeness.
• Non-random techniques — quota (fill set numbers from each category), snowball (one contact leads to the next), volunteer (self-selecting), opportunity (whoever is to hand), purposive (deliberately choosing informative cases). Weaker on representativeness — but often the only way in.

Why inequality research forces non-random sampling. There is no sampling frame for undocumented migrants, rough sleepers, drug dealers or gang members. So researchers use snowball sampling and rely on a gatekeeper — someone inside the group who grants access and vouches for you. William Foote Whyte got into an Italian-American street corner gang (Street Corner Society, 1943) only through 'Doc'; Sudhir Venkatesh (Gang Leader for a Day, 2008) got into a Chicago housing project only through the gang leader 'J.T.'. The catch: the gatekeeper shapes what you see, steering you towards their allies and away from their enemies.

The representativeness disaster: Shere Hite's Hite Report (1976) distributed 100,000 questionnaires and received about 3,000 back — a response rate of roughly 3%. Those who bother to reply to a long, intimate questionnaire are unlikely to be typical, so however large the raw number, the findings cannot be safely generalised.

Sampling a hidden population

Match the sampling technique to its definition

Tap a technique on the left, then its definition on the right.

Ethics and the BSA

The British Sociological Association's Statement of Ethical Practice sets the standards OCR expects you to know:

• Informed consent — participants should know what they are agreeing to, and be free to withdraw.
• Confidentiality and anonymity — protect identities and data.
• Avoiding harm — physical, psychological, legal or reputational, to participants and to the researcher.
• Honesty — deception should be avoided; covert research must be exceptionally justified.
• Vulnerable groups — children and people with limited capacity require extra safeguards.

The case that defines the debate. Laud Humphreys's Tearoom Trade (1970) covertly observed anonymous sexual encounters between men in public toilets, acting as a lookout. He then recorded car registration numbers, traced home addresses through police contacts, and — in disguise — interviewed the men in their homes about their family lives. The data were sociologically valuable and challenged stereotypes. But there was no consent, deep deception, and real risk of catastrophic harm (this was an era when exposure could mean prosecution and ruin). Use it to argue both sides: the knowledge could not have been obtained any other way — but should it have been obtained at all?

Covert observation always sits on this fault line: James Patrick's A Glasgow Gang Observed (1973) was published under a pseudonym years later, because he had witnessed violence and feared reprisal. Covert work buys validity with deception and danger.

The ethics of going covert

Questionnaires and interviews

• Questionnaires (usually closed questions) — cheap, wide reach, standardised and therefore highly reliable; easy to quantify. But: low response rates (Hite), no chance to clarify, and severe researcher imposition. Useless if your population cannot read or has no fixed address — a serious limit in poverty research.
• Structured interviews — a questionnaire read aloud. Higher response rate, still standardised and reliable. But interviewer effect (people manage their impression in front of a real human, especially about income, racism or offending) and social desirability bias distort answers. Michael Young and Peter Willmott's Family and Kinship in East London (1957) drew a random sample from the electoral register and used structured interviews to map kinship networks in Bethnal Green — a model social survey.
• Unstructured interviews — a guided conversation. High validity and rapport; excellent for sensitive experiences of inequality. But time-consuming, unreliable (no two interviews are the same), small samples, and the interviewer's own class, gender or ethnicity may shape what is said.
• Semi-structured interviews — a set of core questions plus freedom to follow up. The pragmatic compromise, and the most common method in real inequality research.

Apply it: asking a low-income single mother to tick an income band gets you reliability; sitting with her for an hour gets you the meaning of going without. OCR rewards students who show they know which trade-off they are making, and why.

Observation and ethnography

Two axes: participant vs non-participant, and overt vs covert.

• Overt participant observation — you join in, and they know who you are. Eileen Barker's The Making of a Moonie (1984) spent six years with the Unification Church using overt observation and in-depth interviews, and concluded that recruits were not brainwashed but were making choices that made sense from inside their situation. Ethical, but risks the Hawthorne effect — people behave differently when watched.
• Covert participant observation — you join in, and they do not know. James Patrick passed as a gang member in Glasgow; Erving Goffman's Asylums (1961) worked inside a Washington mental hospital, producing his account of the total institution and the 'mortification of the self'. Maximum validity, maximum ethical cost and personal risk.
• Non-participant observation — you watch without joining, e.g. sitting at the back of a classroom recording teacher-pupil interaction. Less intrusive but shallower.
• Ethnography — the immersive study of a group's whole way of life over a long period, usually combining observation, conversation and documents. Paul Willis's Learning to Labour (1977) spent time with twelve working-class boys and showed how their own counter-school culture — not lack of ability — steered them into manual labour. Rich and valid, but tiny sample, unrepeatable, and the researcher's interpretation is doing a lot of work.

Why go covert?

Sort the methods and sources

Tap a card, then tap where it belongs.

Statistical data, documents and content analysis

• Official statistics — produced by the state: the Census, crime figures, the Labour Force Survey. Cheap, large-scale, often longitudinal, and usually representative. But they are collected for administrative, not sociological, purposes; the definitions can change under political pressure; and Marxists argue they are ideological — what the state chooses to count reveals what it wants to manage.
• Non-official statistics — from charities, unions, think tanks and businesses (for example the Joseph Rowntree Foundation on poverty). Often fill the gaps the state leaves, but may carry the organisation's agenda.
• The dark figure of crime — official crime statistics record only crime that is reported and recorded. That is why the Crime Survey for England and Wales (a victim survey) and self-report studies exist, and why the three sources disagree.
• Documents — diaries, letters, emails, official reports. John Scott's four criteria are the standard test: authenticity, credibility, representativeness, meaning.
• Content analysis — systematically counting or coding the content of media. The Glasgow University Media Group (Bad News, 1976, and after) analysed television news coverage of industrial disputes and argued it systematically favoured management framings. Quantitative content analysis is reliable and replicable; qualitative or thematic analysis captures meaning but is harder to replicate.

Triangulation and methodological pluralism

• Triangulation — using two or more methods to cross-check the same finding. If a survey and an ethnography point the same way, confidence rises; if they diverge, that divergence is itself a finding.
• Methodological pluralism — using different methods to build up a fuller picture of different aspects of the problem, rather than to check one another.

Why it matters for inequality research: a survey can tell you how many households cannot heat their homes; only interviews and observation can tell you what that does to a family. Eileen Barker combined observation, interviews and questionnaires over six years precisely because no single method could carry the argument alone.

The catch: mixed methods cost more time and money, require more skills, and do not dissolve the underlying theoretical disagreement — a positivist and an interpretivist looking at the same mixed dataset will still argue about which half of it counts.

Designing a study

The PET framework

When OCR asks you to justify or criticise a method, run the three factors — and then apply them to the specific group being researched.

• P — Practical: time, money, funding, access, the size and location of the population, the researcher's own skills and safety.
• E — Ethical: consent, confidentiality, harm, deception, vulnerable participants, the researcher's legal position.
• T — Theoretical: validity, reliability, representativeness, generalisability — and the underlying positivist / interpretivist commitment.

The mark-scheme difference. A middle answer says 'covert observation is unethical'. A top answer says: 'covert observation would be the only realistic way to reach an active county lines network, since no sampling frame exists and members would refuse an overt researcher — but it breaches BSA guidance on consent and honesty, exposes the researcher to real physical and legal risk, and, as Patrick's Glasgow study shows, may force the sociologist to witness violence they cannot report.' Same knowledge, applied.

Component 02 Section A — the big ideas

Press Finish to see your score.