A topic in the Open Knowledge Graph — a free, open map of 15,290 topics and the order to learn them in.
True experiments are the only design that can establish causal relationships because researchers actively manipulate the independent variable and randomly assign participants to conditions. The key elements are: (1) manipulation of the IV, (2) measurement of the DV, (3) control of extraneous variables, and (4) random assignment. Experiments can be between-subjects (different participants in each condition) or within-subjects (same participants in all conditions), each with distinct advantages and trade-offs.
Design a simple experiment from start to finish: write a hypothesis, specify the IV and DV operationally, describe the control condition, and explain the random assignment procedure.
You have already learned what variables are in psychological research and how hypotheses are formed. The experimental design is the structure that lets you test whether one variable actually *causes* changes in another — not just whether they tend to occur together. This distinction between causation and correlation is the central reason experiments exist. Observational studies can detect patterns; only experiments can, under the right conditions, establish that one thing makes another thing happen.
The logic of a true experiment rests on random assignment. If you take a group of participants and randomly allocate them to conditions, then on average, every characteristic that might influence the outcome — prior knowledge, motivation, age, personality — will be distributed evenly across your groups. Any systematic difference you observe at the end is therefore attributable to the one thing that differed between groups: the manipulation you introduced. Without random assignment, this logic breaks down. A convenience sample that self-selects into groups carries all its pre-existing differences into the comparison, making it impossible to isolate the effect of your intervention.
The four elements of a true experiment work together: (1) You manipulate the independent variable (IV) — the presumed cause. (2) You measure the dependent variable (DV) — the presumed effect. (3) You control extraneous variables that could otherwise explain your results. (4) You randomly assign participants to conditions. Remove any one of these and you weaken your ability to draw causal conclusions. A study that manipulates an IV and measures a DV but lacks random assignment is a quasi-experiment — useful, but with greater ambiguity about causation.
The choice between between-subjects and within-subjects designs is a practical trade-off. Between-subjects designs give each participant only one condition, so there is no risk that experiencing one condition influences behavior in another (carryover or practice effects). Within-subjects designs use the same participants across conditions, which dramatically reduces the number of participants you need and removes all individual-difference variability from the comparison — a major statistical advantage. The cost is managing order effects through counterbalancing (varying the sequence of conditions across participants).
One important misconception: laboratory experiments do not automatically have more validity than field experiments. Laboratory settings offer control — you can hold constant many variables that vary in the real world — but that control can come at the cost of ecological validity: the degree to which your findings generalize to natural settings. If you are studying whether noise affects test performance, a lab study gives you clean control over noise levels but may produce a situation so artificial that participants behave differently than they would in a real exam. Matching the setting to the research question is what determines validity, not the setting itself.