Definition · Plain-language
Experimental design
Experimental design is the planned structure of a study in which a researcher manipulates an independent variable under controlled conditions to test its effect on an outcome.
The step most authors miss
Doing CRediT right? Don’t stop at the statement.
A CRediT statement credits you inside one paper. The recognition CRediT was built for happens when those roles are tied to you, persistently. Sign in with your ORCID — free — and claim your CRediT contributions on casrai.org, the home of the standard. They become a verified, portable part of your identity, not a line that disappears into one PDF.
Free: claim your contributions, then export a journal-ready CRediT statement, schema.org structured data, JATS XML, CSV or BibTeX — and preview your public profile. A membership publishes that profile publicly and verifies the journals you serve.
Manipulation, control and randomisation
A true experiment rests on three features. Manipulation means the researcher actively varies the independent variable, creating the conditions to be compared rather than merely observing what occurs. Control means holding other influences constant or accounting for them, so that any difference in the outcome can be traced to the manipulation rather than to extraneous variables. Randomisation — random assignment of participants to conditions — distributes known and unknown confounders evenly across groups by chance. Together these allow strong causal inference, which is what distinguishes a true experiment from observational and correlational research.
Between-subjects and within-subjects designs
Two basic structures arrange how participants meet the conditions. In a between-subjects (independent-groups) design, different participants are assigned to different conditions, so each person experiences only one level of the independent variable; this avoids carry-over effects but needs more participants and risks group differences. In a within-subjects (repeated-measures) design, the same participants experience every condition, which controls for individual differences and is more efficient, but introduces order or practice effects that are managed by counterbalancing. Matched-pairs designs sit between the two, pairing similar participants across conditions to combine some advantages of each.
Factorial designs and validity
A factorial design manipulates two or more independent variables simultaneously, allowing the researcher to examine each variable’s separate (main) effect and, crucially, how the variables interact — an effect invisible to single-variable studies. Whatever the structure, a good experimental design protects both internal validity, the confidence that the manipulation caused the outcome, and external validity, the extent to which results generalise beyond the study. Threats such as confounding, selection effects and demand characteristics are anticipated and controlled at the design stage, because no amount of analysis can fully repair a flawed design after the data are collected.
Key facts
At a glance
- Definition: the structure of a study that manipulates a variable to test cause and effect
- Three pillars: manipulation, control and randomisation
- Between-subjects: different participants in each condition
- Within-subjects: same participants in every condition (repeated measures)
- Factorial: manipulates two or more variables to reveal interactions
- Protects: internal validity (causation) and external validity (generalisation)
Common misconceptions
What people often get wrong
Often heard: Any study that collects data and compares groups is an experiment.
Actually: A true experiment requires the researcher to manipulate the independent variable and randomly assign participants. Comparing pre-existing groups without manipulation is quasi-experimental or observational, not a true experiment.
Often heard: Random assignment and random sampling are the same thing.
Actually: They differ. Random sampling selects who is in the study, protecting external validity; random assignment allocates participants to conditions, protecting internal validity. A study can use one without the other.
Often heard: Within-subjects designs are always better because they need fewer participants.
Actually: They are efficient but introduce order and practice effects that must be counterbalanced. Between-subjects designs avoid carry-over but need more participants; the right choice depends on the research question.
