Glossary
External validity
What is External Validity?
External validity is the extent to which study findings can be generalized to other populations, settings, times, and conditions beyond those specifically examined in the study.
How it works
A study of college students may not answer the same question for older adults. A laboratory task may differ from the real setting. Findings can also depend on when the study took place and how the intervention was delivered. Each difference is a reason to examine whether the result transfers. Internal and external validity answer different questions. A design may face trade-offs, but better control does not necessarily reduce generalizability; Campbell and Stanley identify strength in both as the ideal.
Applied example
Suppose a medication trial includes only men aged 25–45 without other health conditions. The result leaves a question about older women with several conditions. Those differences need to be examined before assuming the treatment works in the same way for that population.
Why it matters
“It worked in a study” is the beginning of an application decision. External validity asks whether the people, setting and delivery are close enough to the intended use—or whether more evidence is needed.
Will an evening-class result transfer?
Consider a hypothetical randomized study in which an evening class improves course completion among volunteers at a city campus. A college with several rural campuses wants to offer the same class. Before applying the estimate, specify the target: all eligible rural students, offered the class during the next academic year, with completion measured at the end of the term.
Now compare the situations. The original students volunteered and lived near public transport. The new students may work night shifts, travel much farther, or have unreliable internet for remote attendance. These are not reasons to assume the intervention fails. They identify plausible differences in access and participation that the original study did not settle. Changing the delivery to remote classes might solve a transport problem while also creating a different intervention.
A useful next study could test that delivery in the target campuses and record the constraints expected to alter its effect. If using the original data to reweight the result toward the target population, the measured characteristics must include the relevant differences and the study must contain people who represent those combinations. A statistical adjustment cannot manufacture evidence about people and circumstances absent from the source study. Degtiar and Rose explain the assumptions behind generalization and transport.
External validity versus ecological validity
Ecological validity usually concerns how tasks, responses and settings relate to everyday conditions. External validity is broader: it asks where a result applies, including different people, times and versions of an intervention. A trial at a real city campus is realistic in that setting; it does not thereby answer the question for every campus. Equally, a laboratory finding may transfer when the feature that produces it is shared with the target setting.
For study planning, write the target population beside the actual recruitment plan. That simple comparison often exposes the largest uncertainty before any data are collected. The research-methods guide shows how to include it in a design brief.