Theoretical Sampling

The relation between data and theory rests at the heart of empirical social science, and SAMPLING helps to define that relationship. RANDOM SAMPLING are used to generate DATA for purposes of verifying hypotheses derived from existing THEORY. In contrast, theoretical sampling, which is contained within the GROUNDED THEORY approach (Glaser & Strauss, 1967), is an analytical process of deciding what data to collect next and where those data should be found. It is a version of ANALYTIC INDUCTION in that an emphasis is placed on the analyst jointly collecting, CODING, and analyzing data for purposes of developing theory as it systematically emerges from data.

The central issue in this procedure centers on the theoretical purposes for collecting new data as well as deciding what kinds of data are needed for those purposes. This process goes hand in hand with the CONSTANT COMPARISON method, in which categories, situations, and analytical dimensions are continuously searched for during the research process. By combining theoretical sampling and constant comparison, theory that is tightly integrated and tied to its supporting data can be developed more efficiently.

Integration of a developing theory can be provided through several means, but emphasis is placed on two primary mechanisms. The first is selecting new data sites that specifically address provisional hypotheses grounded in data. Maines (1992) illustrates this process in his study of New York subway riders. Attempting to explain rider distributions, he first collected observational data from riders in subway cars, which led to a hypothesis suggesting that observational data should be collected from riders' waiting positions on subway platforms, which then led to a new hypothesis suggesting that observational data be collected on riders' routes of movement through the cars, which finally led to a hypothesis requiring interview data comparing riders who remained in the car they entered with those who walked to another car before taking a seat. At each stage of the research project, provisional theories were constructed that explained existing data, and qualities of those provisional theories suggested the kinds and sources of new data required to improve the explanatory capacity of the developing theory.

[Page 1122]

The second mechanism is CODING. Strauss (1987, Chap. 3) identifies three types of coding procedures that are linked to theory development. The first is open coding, in which data are explored for purposes of discovering categories and central themes. This process typically occurs early in the research process. The second, axial coding, focuses on the identification of core categories and their properties, and theorizing the conditions under which these properties (or dimensions) would occur as well as their relationships to other categories and their properties. The third procedure is selective coding, which entails the more theoretically purposeful specification of relationships among categories and subcategories.

Soulliere, Britt, and Maines (2001) demonstrate how combining theoretical sampling and coding for core categories is compatible with conceptual modeling. Focusing on the category of “Professionalism,” they show how a concept-indicator model can be respecified on the basis of new data and presented as a concept-concept model. This modeling process is an explicit application of the grounded theory tenet of moving theory from concrete to more abstract and generic formulations.

...