Generalizability

Generalizability in Quantitative Research

In quantitative research, a study's generalizability is assessed by focusing on a study's sampling procedures and by using statistical analysis designed to determine the likelihood that the study's results might have occurred by chance. If a study's sample is large enough and was selected randomly, the statistical analysis is likely to show that there is only a very slim possibility that the study's results are a product of chance. In such cases, the study is said to exhibit external validity and the study's findings are judged generalizable.

For a variety of reasons, including the complexity of social phenomena and the ever-changing cultural dimension of social life, quantitative researchers have had difficulty producing even probabilistic findings that can be generalized even to delimited populations. In the 1970s, this failure led one prominent quantitative researcher, Lee Cronbach, to suggest that quantitative researchers should add a qualitative component to experimental studies so that qualitative researchers' thick descriptions could be used ex post facto to generate grounded hypotheses about why idiosyncratic results occurred. By the 1980s, Cronbach went even further: He argued that action in the social world was constructed, not caused, and, consequently, that those who were waiting for social scientists to produce generalizable findings were, in essence, waiting for Godot.

Generalizability in Qualitative Research

Cronbach's position in the 1980s is quite similar to the position articulated by qualitative researchers from the symbolic interactionist, ethnomethodologist, and other constructivist traditions. But, in fact, virtually all qualitative researchers have been forced to rethink quantitative researchers' generalizability notion, if only because of their small samples.

Because of small sample sizes, most qualitative researchers simply cannot play the traditional general-izability game. Instead, they have redefined the generalizability question in more commonsense terms: Why will knowledge of a single or limited number of cases be useful to people who operate in other, potentially different situations?

Qualitative researchers have answered this question in a number of ways. Yvonna Lincoln and Egon Guba, for instance, took note of the same sort of complexity that Cronbach had discussed and concluded that only consumers of research could determine whether the setting studied was sufficiently similar to the consumer's organization to entertain the working hypothesis that findings would transfer. Robert Donmoyer, on the other hand, suggested that reading qualitative accounts of radically different cases could produce enriched cognitive schema and that these schema would allow for a kind of intellectual generalization even when settings are radically different.