Revise Research methods in psychology for IB Psychology (4.1)

The experimental method is the only method that can establish cause and effect. The researcher manipulates one variable — the independent variable (IV) — and measures its effect on another — the dependent variable (DV) — while controlling all other relevant variables. Both variables must be operationalised: defined in specific, measurable terms. 'Memory' is not measurable, but 'the number of words correctly recalled from a 20-item list in two minutes' is.

Control is what makes the causal claim possible. If everything except the IV is held constant, then any change in the DV must be due to the IV rather than to some other factor. An uncontrolled factor that varies with the IV is a confounding variable, and it offers a rival explanation that undermines the study. The whole design of an experiment is really an effort to rule out these rival explanations.

Experiments come in three types that trade control against realism. A laboratory experiment offers the highest control in an artificial setting, maximising internal validity but often at the cost of ecological validity. A field experiment manipulates the IV in a real-world setting, gaining ecological validity but sacrificing some control. A natural experiment studies a naturally occurring IV that the researcher cannot manipulate (such as the effect of a disaster or a policy change); because the IV is not truly manipulated, causal claims are weaker and confounds harder to exclude.

A correlational study measures two variables that already exist and examines whether they are related — it manipulates nothing. It can tell you the direction (positive or negative) and strength of a relationship, which is valuable when manipulation would be impractical or unethical. But it has a hard limit: a correlation can never, on its own, establish cause and effect. Two rival explanations always remain. First, the bidirectionality problem: even if A and B are related, A might cause B or B might cause A. Second, the third-variable problem: some other factor may cause both. A famous illustration is that ice-cream sales correlate with drowning — but neither causes the other; hot weather (the third variable) drives both.

A case study is an in-depth, often longitudinal investigation of a single individual, group or event, usually drawing on several sources of data (interviews, observations, records). Its great strength is depth and holism — it can explore rare or complex cases (such as a patient with unusual brain damage) in a richness no experiment could match. Its limitations follow directly from its method: findings from a single case are difficult to generalise, and the close involvement of the researcher can introduce bias.

Observations record behaviour as it occurs and vary along two dimensions. A naturalistic observation watches behaviour in its normal setting; in a participant observation the researcher joins the group being studied. Independently, observation can be overt (participants know they are being watched, which risks reactivity — people behaving differently because they are observed) or covert (they do not know, which improves validity but raises consent and privacy concerns). As with every method, the choice shapes the conclusions: a covert naturalistic observation captures genuine behaviour but cannot easily explain why it happened.

Data come in two broad kinds, and method choice usually determines which. Quantitative data are numerical and can be statistically analysed — reaction times, error counts, questionnaire scores. They allow objective comparison and are typical of experiments and correlational studies. Qualitative data are non-numerical and descriptive, capturing meaning and experience — interview transcripts, field notes, open-ended responses. They offer depth and context and are typical of case studies and many observations. Neither is superior; they answer different questions. Quantitative data are good for measuring how much or how many; qualitative data are good for understanding why and how something is experienced.

A sample is the group actually studied; the target population is the wider group the researcher wants to describe. Generalisability is the extent to which findings from the sample can be applied to that population, and it depends heavily on how the sample was drawn. Even a flawlessly conducted study tells you little about the wider world if its sample is biased.

These two ideas are the criteria by which any study is judged, and confusing them is a classic error. Validity is about accuracy — does the study measure what it claims to measure? Reliability is about consistency — would repeating the study produce the same results? A study can be reliable but not valid: a scale that always reads 3 kg too heavy is perfectly consistent yet always wrong.

Every strand of this lesson points to a single principle: the method sets a ceiling on what a study can legitimately claim. An experiment can claim causation but may lack realism; a correlation can claim a relationship but never a cause; a case study offers depth but not breadth; an observation captures real behaviour but not its reasons. A strong evaluation always ties its point back to the method — not 'the study is weak', but 'because this was a correlational study, it cannot show that X caused Y, only that they are related'.