Psychology research

Core idea

Psychology earns the label "science" by following a disciplined loop: state a hypothesis that two variables are related, manipulate one of them under controlled conditions, measure the other, and analyse whether the difference is bigger than chance would produce. Everything that follows in the field — the credibility of a result, the strength of a claim, the size of the effect — depends on how cleanly each step in this loop was executed.

Two clusters of concepts govern the loop. The first is design — independent and dependent variables, control groups, random assignment, the choice between an experiment and a correlation, the choice between observing and intervening. The second is threats — extraneous variables, demand characteristics, observer bias, the Hawthorne effect, the gap between the laboratory and the world. Good research is the constant management of those threats.

Why it matters

When you understand the loop, you can read a psychology finding the way a forensic accountant reads a balance sheet: not as a verdict, but as a set of claims whose support you can audit.

Mental model

The research loop

The loop runs from a question through measurement to a published finding, and then back into the question via replication and revision.

Copy sourceZoom

Experiment versus correlation

The single biggest source of confusion in popular reports of psychology is treating a correlation as if it were an experiment. They answer different questions.

Copy sourceZoom

Two kinds of validity

Internal validity is about the inside of the study; external validity is about the outside of it. A study can be high on one and low on the other.

Copy sourceZoom

Practical application

Example

Imagine a headline: "Workers in offices with plants are 38 per cent more productive." You should treat that as a hypothesis report until you have done the audit.

Look for the design. If the study compared two offices — one with plants, one without — and the workers were assigned at random to each, you have a quasi-experiment with reasonable internal validity. If the comparison was between offices that already had plants (because their managers liked plants) and offices that did not, you have a correlation. The two are not equivalent. In the correlational case, the managers who choose to install plants might also pay more, manage better, or hire differently — and any of those could explain the productivity gap.

Now look for the threats. If workers knew they were being observed (a likely Hawthorne effect), productivity might rise simply from the attention. If productivity was measured by self-report rather than by output, demand characteristics could inflate it. If the study ran for two weeks, a maturation threat could matter — workers might just be getting better at their jobs.

You are not dismissing the finding. You are pricing it. The 38 per cent number is probably an overestimate of the causal effect, but the direction is plausible. The next question — and this is the discipline of psychology working as intended — is what a tighter replication, with random assignment and an objective output measure, would show.

Related lessons