Analytic Induction

Analytic induction is sometimes treated as a model for qualitative research design. It portrays inquiry as an iterative process that goes through the following stages: initial definition of the phenomenon to be explained; examination of a few cases; formulation of a hypothesis; study of further cases to test this hypothesis; and, if the evidence negates the hypothesis, either redefinition of the phenomenon to exclude the negative case or reformulation of the hypothesis [Page 17]so that it can explain it. This is followed by study of further cases, and if this further evidence supports the hypothesis, inquiry can be concluded. However, if any negative case arises in the future, inquiry will have to be resumed: Once again, either the hypothesis must be reformulated or the phenomenon redefined, and further cases must be studied, and so on.

Analytic induction shares with grounded theory an opposition to the testing of hypotheses derived from armchair theorizing (see induction). It also shares with that approach an emphasis on the development of theory through investigating a relatively small number of cases and adopting a flexible mode of operation in which theory emerges out of—and guides future—data collection and analysis. At the same time, analytic induction has features that are at odds with the self-understanding of much qualitative research today—notably, its explicit concern with testing hypotheses and discovering universal laws. Its most distinctive feature is recognition that the type of phenomenon to be explained may need to be reformulated in the course of research; this arises because the task is to discover theoretical categories all of whose instances are explained by a single type of cause. In the context of analytic induction, a theory specifies both necessary and sufficient conditions, and this is why a single exception is sufficient to refute any theory.

The term analytic induction seems to have been coined by the Polish American sociologist Florian Znaniecki in his book The Method of Sociology (Znaniecki, 1934). There, he contrasted it with enumerative induction, a term referring to the kind of inference that is characteristic of statistical method. Enumerative induction involves trying to discover the characteristics of phenomena belonging to a particular class by studying a large number of cases and describing what they have in common. Znaniecki points to a paradox here: If we do not know the essential features of a type of phenomenon, we cannot identify instances of it; however, if we do know those essential features, we already know what enumerative induction is designed to tell us. Following on from this, Znaniecki argued that, contrary to the claims of some contemporary social scientists, statistical method is not the method of natural science: Natural scientists do not study large numbers of cases and try to derive theories from their average characteristics. Rather, he suggested, they study a single case or a small number of cases in depth, and develop theories that identify the essential features of these cases as instances of general classes of phenomena. Thus, according to Znaniecki, the task of science is the search for universal laws—relations of causality or functional dependence—not statistical generalizations; and it is analytic, not enumerative, induction that is its method (see Gomm, Hammersley, & Foster, 2000; Hammersley, 1989).

...