External Validity

External ValidityContrasted With Internal Validity

Many textbooks and articles on research design published before 1957 discuss how the sampling techniques used to gather observations will affect the generalizability of research findings. After 1957, it is much more common to find generalizability [Page 655]discussed in terms of the external validity of a research design. This change in terminology, and the increase in attention given to questions of generalizability, followed publication in 1957 of an article by the psychologist and methodologist Donald Campbell in which Campbell reframed the generalizability question as one of “external validity” to be contrasted with questions of “internal validity.” In this 1957 article and a series of subsequent influential publications, Campbell and coauthors examined threats to external validity and ways of increasing and assessing external validity. Although many scholars have made important contributions on the topic of external validity, Campbell and his colleagues’ work on this topic continues to serve as the foundation for most other discussions.

The primary insight of Campbell, and his reason for placing internal and external validity in contrast, was that steps taken in the research design process to increase internal validity often decrease external validity, and steps taken to increase external validity often decrease internal validity. Thus, if we limit who may participate in a study to reduce the chance that individual-level differences will confound the result (resulting in greater internal validity), we cannot be sure the results will generalize to other groups of persons (resulting in less external validity). A compromise would be to study a more representative sample drawn from the population about which one wants to draw inferences while measuring the individual difference-level variables that the researcher has reason to believe may affect the nature or degree of relationship between the target variables (e.g., the researcher may record the sex of students who serve as participants in a study on the relation of teaching style to course grades in order to examine whether male and female students exhibit the same pattern of results).

...