Beneath every well-designed study lies a set of logical building blocks that convert a curiosity about the social world into a testable, defensible enquiry. Variables specify what is being studied, hypotheses state what is expected, and sampling determines whom we study and with what right to generalise. Reliability and validity then supply the standards by which the resulting evidence is judged. Mastering this vocabulary is indispensable, because Paper I questions on methodology reward precision of terms as much as breadth of illustration.
Variables
A variable is any characteristic that can take different values across cases — age, income, religiosity, crime rate. Research is largely the study of relationships between them.
- The independent variable is the presumed cause, the factor the researcher manipulates or treats as prior.
- The dependent variable is the presumed effect, whose variation we seek to explain.
- The intervening (or mediating) variable stands between the two, transmitting the influence — for example, poverty (independent) may raise ill health (dependent) through poor nutrition (intervening).
Careful researchers also watch for extraneous and confounding variables that create spurious correlations, and they insist on operationalisation — translating an abstract concept such as social class into a measurable indicator such as occupation.
The hypothesis
A hypothesis is a clear, testable statement predicting a relationship between variables — for instance, that educational attainment rises with parental income. A good hypothesis is specific, falsifiable in Popper's sense, and derived from theory. It is often paired with a null hypothesis asserting no relationship, which the data attempt to disprove. Interpretivists, however, frequently reject hypothesis-testing altogether, preferring to let concepts emerge inductively from the field through grounded theory, since imposing prior hypotheses may blind the researcher to the actors' own meanings.
Sampling
Because studying an entire population is rarely feasible, researchers study a sample and infer back. A good sampling frame — a complete list of the population — underpins the whole exercise.
Probability (random) sampling gives every unit a known, non-zero chance of selection, permitting statistical generalisation.
- Simple random sampling — every member has an equal chance, as with a lottery; unbiased but demanding a full frame.
- Systematic sampling — selecting every nth case from a list; efficient, though a hidden periodicity in the list can distort it.
- Stratified sampling — dividing the population into strata (say, by age or sex) and sampling each in proportion, improving representativeness for known variables.
- Cluster (multi-stage) sampling — sampling whole groups such as schools or districts, then units within them; economical over dispersed populations but less precise.
Non-probability sampling does not rest on random selection and cannot support statistical generalisation, yet it is often unavoidable. It includes quota sampling (filling preset category targets), purposive sampling (choosing information-rich cases), snowball sampling (respondents recruiting others, useful for hidden or deviant groups), and convenience sampling. Interpretivist and small-scale qualitative studies rely on these because their aim is depth and theoretical insight rather than representativeness.
Reliability
Reliability refers to consistency — the degree to which a method, repeated by the same or another researcher under the same conditions, yields the same results. It is the hallmark of positivist, quantitative techniques: standardised questionnaires and structured interviews are highly reliable because they are replicable. Reliability answers the question, would we get this result again?
Validity
Validity refers to truth — the degree to which a measure actually captures the reality it claims to capture. A method is valid if it produces a genuine, authentic picture of social life. Qualitative techniques such as participant observation and unstructured interviews are typically strong on validity because they access meaning in context, answering the question, are we really measuring what we think we are?
The tension between them
The two criteria frequently pull in opposite directions. Highly reliable methods achieve consistency by standardising and constraining, which can strip away context and yield superficial, invalid data — official crime statistics are reliable yet may misrepresent real offending. Highly valid methods capture rich meaning but, being unique to their setting and researcher, are hard to replicate and thus unreliable. Positivists prioritise reliability, interpretivists validity; the pragmatic resolution is triangulation, combining methods so that the strengths of one offset the weaknesses of the other. The broader positivism–interpretivism debate is developed in research methods in sociology.
How to use this in the exam
Define each term crisply, then show the relationships between them — this is what separates high scorers. Draw the reliability–validity trade-off explicitly and anchor it in a concrete example such as official statistics versus ethnography. When a question concerns sampling, distinguish probability from non-probability logic and link the choice to the researcher's theoretical stance. Finish by invoking triangulation as the mature response to the tension, demonstrating that you grasp methodology as a coherent argument rather than a glossary.