Research Methods in Psychology: Tests, Interviews, and Observation

Scientific psychology does not just watch everyday life unfold: it gathers data with tools built for the purpose, and each one plays by its own rules. To see how a study ends up saying something about how we think or behave, as explained in psychological science: what it is and how it works, it helps to start with the three methods a researcher has on hand for collecting data about a psychological phenomenon: the test, the interview, and observation.

The psychological test

A psychological test asks everyone the same questions or the same tasks, in the same way, to measure a trait. It is judged on two separate properties, reliability and validity, and mixing them up is the most common mistake.

Reliability is about the consistency of a measure: if the test is repeated, does it give similar results? One specific form of reliability is internal consistency, meaning how well a person’s answers hold together across the different items of the same test; the usual statistic for it is Cronbach’s alpha. Validity is a different question, about content: does the test actually measure what it claims to measure, or something else entirely? One way to check it is criterion validity, which compares the test’s scores against other external measures linked to the same trait.

That is why, in psychology, the two properties are checked separately even when a test seems to work well in practice. The same principle applies to intelligence tests, covered in human intelligence: theories and tests: a score that stays stable over time still needs an independent check before anyone can say the test is capturing the right ability.

Practical example: someone fills out an anxiety questionnaire twice, a month apart, and gets similar scores: that makes it reliable. The problem shows up when those scores have no link at all to other signs of that person’s anxiety, say the accounts given by people close to them: in that case the questionnaire fails the test of validity, because the score has no external backing.

The interview

The interview is the most widely used data-collection method in the social sciences, psychology included. It is defined as a conversation between two people, verbal and non-verbal, with a research purpose set in advance that sets it apart from ordinary small talk.

The three types of interview differ in how much the questions are already decided before the conversation even starts. In an unstructured interview, the interviewer follows topics in whatever order they come up during the conversation, with no fixed script. In a semi-structured interview, the topics are chosen in advance and the interviewer adapts the wording and order of the questions to the person in front of them. A structured interview instead uses an identical questionnaire for everyone, the same questions in the same order for every respondent.

Every type carries the same underlying limit. The same question can be understood differently by different people, which introduces subjectivity into the answers that the researcher does not fully control. There is also the interviewer effect: how the person asking questions behaves calls up roles and habits the respondent has already learned in other social settings, and that can shape what they say.

Observation

Watching looks like the simplest method of all: just look, and no one has to answer a question. In psychology, though, it comes with forms and limits just as precise as the other two tools.

Naturalistic observation follows behavior within its usual setting without the researcher stepping in to change it. A case study instead looks deeply at a single person, or a small group: the level of detail it reaches has no equal, precisely because it focuses on so few cases, which do not represent the wider population.

Whoever is observing, though, is part of the research being carried out, and that alone can be enough to unintentionally change what gets recorded: it is a limit on the side of the watcher, not just the watched. To keep it in check, the same event is often followed by several observers, and researchers then check how closely their notes match: this is inter-observer reliability.

There is a second limit, this time on the side of the people being observed: simply knowing you are being watched can be enough to make people act differently from usual. In the Hawthorne studies, carried out in a factory between 1924 and 1927 and remembered as the Illumination Experiment, productivity rose with almost every change in lighting, whether brighter or dimmer, and stayed high even when conditions worsened again. Researchers attributed the rise to the workers’ awareness of being observed.

Reliability and validity: two different things

Tests, interviews, and observation share this much: the data they produce can describe relationships between variables, for instance that two things tend to occur together or that a behavior recurs in certain situations. Saying that one causes the other calls for a different research design.

This holds no matter how well the tool is built. A reliable and valid test gives sturdier results; the same holds when an interview is carefully conducted or when several people confirm the same observation. The question of why a phenomenon happens still remains, in most cases, territory that needs other research tools.