For decades, political scientists have been taught that it is dangerous if not forbidden to search on the dependent variable. By looking only at cases in which the effect of interest occurs, we cannot infer causation because the factors that we think are powerful may be only necessary conditions, which means that they can be present in cases in which the effect does not appear. They then cannot readily be labelled as causes in the sense of distinguishing between instances in which the effect is present and those in which it is absent. Yet almost all studies of intelligence failures have looked only at the failures themselves, without making comparisons with successes. This was true even of scholars who not only knew about this problem, but had stressed it in their teaching. I can say that with some confidence because I am in that category. While I urged others to look at successes as well as failures, I did not do so.[1] Erik Dahl, one of our reviewers, recently used this methodology,[2] but our reviewers agree that until this volume, no one had sought such careful and controlled comparisons that are developed here.Continue reading