The Logic of Social Research (review)

In his 1968 book, Constructing Social Theories, Arthur Stinchcombe delivered terrific insights on how sociologists ought to think about and study cause and effect. Among other accomplishments, the book distinguished the key elements in theoretical claims; delineated and explicated the logic of different kinds of causal claims, notably demographic, historical and functional explanations; and laid out the scientifically necessary steps needed to make a convincing empirical statement. Constructing Social Theories well deserved its 1970 A.S.A. Sorokin award.

In this 2005 book, The Logic of Social Research, Stinchcombe delivers more insights, this time about methods – the “various strategies sociologists have invented to explore... or test theories of causation”(1) – along with his distinctive complement of digressions ranging from informative and apropos to informative and distracting. An overarching and welcome feature of the book is the way Stinchcombe dismisses tired old divisions among methods, invites the use of many styles of research, from hard statistics to soft interpretations, and shows that deep down they are logically parallel. Also refreshing, Stinchcombe focuses on the practicalities of doing research well rather than on philosophical abstractions about research. (Who needs, for example, yet another stale and barren debate about “positivism”?)

Although billed as a textbook (183), The Logic of Social Research would be difficult to use that way, even for teaching entering graduate students. It presumes that the reader already knows a lot, for example, about partial regression coefficients (4), social constructivism (196), LISREL (257), and who Jessie Helms was (216). Also, Stinchcombe provides little in the way of how-to-do-it recipes. There are general rules of thumb – e.g., “Investigators should regard their problem of having a single case or a few cases to observe as demanding studying mechanisms within the cases...” – but the student facing a fieldwork exercise would need considerable help in translating such rules into practice. And many of Stinchcombe’s extended examples are so subtle, complex or esoteric that they would probably baffle rather than enlighten most students.

The Logic of Social Research works better as a source of ideas for instructors of methods classes – ideas to use or to debate. There is great counsel in the injunction to “‘borrow whatever works to build a theory for wherever one needs it’ [and]... the same advice can be applied to methods.”(301) Also, Stinchcombe’s catalog of “social mechanisms” (Chapter 6) is a useful heuristic guide for spotting causal processes; the “bargain,” one such mechanism, connects pieces of many different social systems together, such as the state and business, and can be a fruitful focus of study. As another example, Stinchcombe’s combined discussion of interaction effects, boundaries and scope conditions (279ff) might be used to help explain the conditionality of empirical claims.

Yet other treatments in the book might stir musings and even useful debate. For example, Stinchcombe relies heavily on regression as the archetype of causal analysis and on the partial regression coefficient as the exemplary proof of causality. But he does not address critiques (by Lieberson and by Abbott, for example) of sociologists’ reliance on the logic of regression, nor does he explain how regression analysis is, ultimately, a crude effort to simulate the more powerful causal logic of experiments. Stinchcombe repeatedly advises researchers to sample at the extremes in order to maximize variances in presumed causes and effects. This is often great advice, but sometimes, if the association between variables is not linear or not monotonic, extremes can mislead us about the processes in between. An S-shaped diffusion function, for instance, cannot be identified from the extremes. (Indeed, much, perhaps most, actual social research is devoted to just describing the full distributions of individual variables, for example, of income, family size or crime victimization.) And Stinchcombe notes a couple of times that “sociologists measure everything badly” (292), but does not devote the space to handling that problem that is commensurate with its gravity.

Art Stinchcombe is a smart and stimulating author (even if he sometimes stimulates the reader with...