For nearly a century, investigators in the social sciences have used regression models to deduce cause-and-effect relationships from patterns of association. Path models and automated search procedures are more recent developments. In my view, this enterprise has not been successful. The models tend to neglect the difficulties in establishing causal relations, and the mathematical complexities tend to obscure rather than clarify the assumptions on which the analysis is based. Formal statistical inference is, by its nature, conditional. If maintained hypotheses A, B, C,... hold, then H can be tested against the data. However, if A, B, C,... remain in doubt, so must inferences about H. Careful scrutiny of maintained hypotheses should therefore be a critical part of empirical work-- a principle honored more often in the breach than the observance.

* This paper will be published in Gerardo L. Munck and Richard Snyder, Passion, Craft, and Method in Comparative Politics (Baltimore, MD: Johns Hopkins University Press, forthcoming, 2007). I would like to acknowledge the helpful comments I received from Robert Adcock, Andrew Gould, Richard Snyder and one anonymous reviewer. This paper focuses on the past and present of comparative politics in the US. The discussion is organized around three issues: the definition of the field’s subject matter, the role of theory, and the use of methods. These three issues are the basis for an identification of distinct periods in the history of comparative politics and for assessments of the state of the field. Attention is also given to the link between comparative politics, on the one hand, and other fields of political science and other social sciences, on the other hand, and, more briefly, to political events and the values held by scholars of comparative politics. The evolution of comparative politics is seen as punctuated by two revolutions: the behavioral revolution, during the immediate post-World War II years until the mid-1960s, and the second scientific revolution, which started around the end of the Cold War and is still ongoing. On both occasions, the impetus for change came from developments in the field of American politics and

s huge. Our purpose in providing this document, a revised version of a portion of a grant proposal submitted to the National Science Foundation in 1996, is two-fold. One is that it provides a (necessarily selective and incomplete) documentation of some of the impressive advances in the study of elections, public opinion, and related areas that have been based on NES data. Given the central role these data have played in these fields, this document should constitute a helpful review of a crucial part of the relevant literatures. Second, this document should also help stimulate ideas about further research that might tap some of the vast unused potential of these 2 individual studies, as well as the "greater-than-the-sum-of-the-parts" potential of the extensive time series. In the end, the value of the project depends on the ideas and work not just of those who design the study, but those whose inspiration leads to research based on the data. All who venture within are

Although the use of models has come to dominate much of the scientific study of politics, our understanding of the role or function that models play in the scientific enterprise has not kept pace. Political science clings to an outdated theorybased approach to scientific inference known as hypotheticodeductivism and is now passing it on to a new generation of scholars through the EITM project. We argue for a new approach to scientific inference that highlights the centrality of models in scientific reasoning, avoids the pitfalls of the hypothetico-deductive method, and offers political scientists a new way of thinking about the relationship between the models with which we are so familiar and the natural world.