AP® Psychology · Science Practices: Research Methods and Data
2.A Experimental and Non-Experimental Designs
An experiment is the only design that can support a cause claim, because it manipulates a variable and randomly assigns participants to conditions, ruling out other explanations. Correlational and case study designs can describe or explore relationships but never prove what caused what.
Research designs split into two families: experimental and non-experimental. In an experiment, the researcher manipulates an independent variable, assigning participants to different conditions, and then randomly assigns who goes where. Random assignment is the key move: it spreads differences in motivation, ability, personality, and everything else roughly evenly across groups, so that if the groups end up different on the outcome, the manipulated variable is the most believable explanation left standing. That is why only an experiment earns the word caused.
Non-experimental designs skip one or both of those steps. A correlational study measures two variables as they naturally occur, without manipulating anything, and can only report that they relate, not why. A case study digs deeply into one person or small group, generating rich description and hypotheses but no basis for generalizing or claiming causation at all. Picture the same question, does sleep affect memory, studied three ways: an experiment randomly assigns people to sleep 4 or 8 hours and compares recall the next day; a correlational study asks people how much they slept and how well they remember, and just checks whether the two numbers move together; a case study follows one insomniac in detail over months. Only the first can say sleep caused the memory difference.
This distinction sets up everything else in research methods: how confounds sneak in, why ethics sometimes blocks manipulation entirely, and how to read a study's design before trusting its conclusion. When a variable cannot be manipulated for ethical or practical reasons, like studying the effect of childhood abuse, researchers rely on correlational or quasi-experimental designs and must speak only in terms of association or risk, never causation.
Key terms
- Manipulation
- The researcher directly controls or changes the independent variable rather than just observing it.
- Random assignment
- Assigning participants to conditions by chance so groups start out roughly equivalent on all other traits.
- Confound
- An outside variable that differs systematically between groups and offers an alternative explanation for the results.
- Correlational study
- A design that measures two variables as they naturally occur and can show they relate but not that one causes the other.
- Case study
- An in-depth examination of one individual or small group, useful for description but not for causal or generalizable conclusions.
Watch it explained
- How it worksHow does manipulating a variable and assigning at random separate an experiment from everything else?
- How it worksWalk me through the same question studied as an experiment, a correlation, and a case study.
- Common mistakeWhy isn't a study with a control group automatically an experiment?
- Going furtherWhat changes about the conclusions when the variable cannot ethically be manipulated?
Check yourself
End of topic.