• Choose a category
• All
• Classic-Fiction

By Root 802 0

case studies should be distinguished from the deductive development of theory. Deductive methods can usefully develop entirely new theories or fill the gaps in existing theories; case studies can test deductive theories and suggest new variables that need to be incorporated. (The literature on deterrence, as noted below, provides an excellent example of this process.) But theory development via case studies is primarily an inductive process. This section highlights the usefulness of deviant cases for inductively identifying new variables or causal mechanisms. (Plausibility probes, which we do not discuss here, also focus directly on the goal of theory development, by aiming at clearer specification of a theory and its variables and by attempting to better identify which cases might prove most valuable for theory building.)

THEORY DEVELOPMENT AND HISTORICAL EXPLANATION OF SINGLE CASES

The outcome in a deviant case may prove to have been caused by variables that had been previously overlooked but whose effects are well known from other research. This leads to an improved historical explanation of the case, but not necessarily to any new generalizations from the case, unless the case is one in which the previously overlooked variables were not expected to have any effect.

An inductively derived explanation of a case can also involve more novel theories and variables. In this context, researchers are frequently advised not to develop a theory from evidence and then test it against the same evidence; facts cannot test or contradict a theory that is constructed around them. In addition, using the same evidence to create and test a theory also exacerbates risks of confirmation bias, a cognitive bias toward affirming one’s own theories that has been well documented both in laboratory experiments and in the practices of social scientists.218

However, it is valid to develop a theory from a case and then test the theory against additional evidence from the case that was not used to derive the theory. This makes the theory falsifiable as an explanation for the case, and can circumvent confirmation biases. Researchers, even when they are fairly expert on a case and its outcome (or the value of its dependent variable), are often ignorant of the detailed processes through which the outcome arose.219 As a researcher begins to delve into primary sources, there are many opportunities to reformulate initial explanations of a case in ways that accommodate new evidence and also predict what the researchers should find in evidence they have not yet explored or had not even thought to look for. Researchers can also predict what evidence they should find in archives before these are made accessible or in interviews before they are carried out.220 Indeed, in testing a historical explanation of a case, the most convincing procedure is often to develop an explanation from data in the case and then test it against other evidence in the case; otherwise, the only recourse is to test the explanation in other cases that differ in ways that may prevent generalization back to the original case.

THEORY DEVELOPMENT AND CONTINGENT GENERALIZATIONS

The study of a deviant case can lead a researcher to identify a new type of case. As we discuss in Chapter 11, this process can take place through a “building block” approach, with new case studies identifying subtypes or the causal processes that apply to a subtype of cases. Each case study thus contributes to the cumulative refinement of contingent generalizations on the conditions under which particular causal paths occur, and fills out the cells or types of a more comprehensive theory.

Historians often view efforts to generalize from historical case studies with suspicion. Yet one can generalize from unique cases by treating them as members of a class or type of phenomenon; that is, as instances of alliance formation, deterrence, war initiation, negotiation, peacekeeping, war termination, revolution, and so on. This is often followed by distinguishing subclasses of each of the phenomena. Researchers