¹ George, Alexander L. and Smoke, Richard, Deterrence in American Foreign Policy York:Columbia University Press, 1974Google Scholar); Jervis, Robert, The Logic of Images in International Relations (Princeton:Princeton University Press, 1970Google Scholar); Jervis, , Perception and Misperception International Politics (Princeton: Princeton University Press, 1976Google Scholar); Ned Lebow, Richard, Be tween Peace and War: The Nature of International Crisis (Baltimore:The Johns Hopkins University Press, 1981Google Scholar), and “Windows of Opportunity: Do States Jump through Them?” International Security 9 (Summer 1984), 147CrossRefGoogle Scholar–86; Snyder, Glenn H. and Diesing, Paul, Conflict among Nations: Bargaining, Decision Maying, and System Structure in International Crises (Princeton:Princeton University Press, 1977Google Scholar); Steinbruner, John D., The Cybernetic Theory of Decision (Princeton University Press, 1974Google Scholar). Also see the following chapters in Jervis, Robert, Ned Lebow, Richard, and Gross Stein, Janice, eds., Psychology and Deterrence (Baltimore: The Johns Hopkins University Press, 1985Google Scholar): Jervis, “Introduction and Assumptions,” 1–12, and “Perceiving and Coping with Threat,” 13–33; Stein, “Calculation, Miscalculation, and Conventional Deterrence I: The View from Cairo,” 34–59, and “Calculation, Miscalculation, and Conventional Deterrence II: The View from Jerusalem,” 60–88.

Many other excellent recent case studies in strategic affairs are excluded from this list, either because they do not deal directly with the theory of rational deterrence (e.g., Posen, Barry R., The Source of Military Doctrine [Ithaca, NY:Cornell University Press, 1984Google Scholar]; Betts, Richard K., Nuclear Blackmail and Nuclear Balance |Washington, DC:Brookings, 1987Google ScholarI; Walt, Stephen M., The Origins of Alliances [Ithaca, NY:Cornell University Press, 1987Google ScholarI), or because they do not criticize it (Mearsheimer, John, Conventional Deterrence [Ithaca, NY:Cornell University Press, 1983Google Scholar]). In addition, we pay relatively little attention to two other important discussions: Snyder and Diesing (fn. 1), which has been discussed elsewhere (see Duncan Sni-dal, “The Game Theory of International Politics,” World Politics 38 [October 1985], 25–57); and Lebow, Richard Ned and Stein, Janice Gross, “Beyond Deterrence-Building Better Theory,” Journal ofSocial Issues 43 (1987), 5–71CrossRefGoogle Scholar, which appeared after this article was written.

² George and Smoke (fn. 1), 503.

³ Lebow, Between Peace and War (fn. 1), 274.

⁴ Richard Ned Lebow, “Conclusions,” in Jervis, Lebow, and Stein (fn. 1), 203.

⁵ The charge is not that deterrence theory is devoid of any appreciation of empirical materials; the major works (e.g., Schelling, Thomas C., Arms and Influence [Cambridge:Harvard University Press, 1966Google Scholar]) are replete with examples from across the millennia. Instead, the concern is with the well-known dangers of facile rational reconstruction of events. Intensive case studies-not attempted by scholars such as Schelling-provide a check.

⁶ Stein, “The View from Cairo” (fn. 1), 34; see also George, Alexander L., “Case Studies and Theory Development: The Method of Structured, Focused Comparison,” in Lauren, Paul Gordon, ed., Diplomacy: New Approaches in History, Theory and Policy (New York:Free Press, 1979), 48Google Scholar.

⁷ Lebow, Between Peace and War (fn. 1), 6.

⁸ This sort of transition in thinking is hardly unique to deterrence analyses. It parallels the general shift in the understanding of the methodology of comparative analysis throughout political science. For example, George's discussion of the problem (fn. 6) acknowledges significant debts to Eckstein, Harry, “Case Study and Theory in Political Science,” in Greenstein, Fred I. and Polsby, Nelson W., eds., Handbook of Political Science (Reading, MA:Addison-Wesley, 1975), Vol. 7, pp. 79–138Google Scholar; also see Lijphart, Arend, “Comparative Politics and the Comparative Method,” American Political Science Review 65 (September 1971), 682CrossRefGoogle Scholar–93, and Lijphart, , “The Comparable-Case Strategy in Comparative Research,” Comparative Political Science 8 (July 1975), 158CrossRefGoogle Scholar–77.

⁹ George and Smoke (fn. 1), chaps. 4 and 16; George (fn. 6); and George, , “Case Studies and Theory Development,” presented to the Second Annual Symposium on Information Processing in Organizations, Carnegie-Mellon University, Pittsburgh, October 15–16, 1982Google Scholar; George, Alexander L. and McKeown, Timothy J., “Case Studies and Theories of Organizational Decision Making,” Advances in Informational Processing in Organizations 2 (1985), 21–58Google Scholar.

¹⁰ George (fn. 6).

¹¹ Janis, Irving L., Victims of Groupthink (Boston:Houghton Mifflin, 1972Google Scholar).

¹² Lebow, Between Peace and War (fn. i).

¹³ Steinbruner (fn. i), and Steinbruner, , “Beyond Rational Deterrence: The Struggle for New Conceptions,” in Knorr, Klaus, ed., Power, Strategy, and Security (Princeton:Princeton University Press), 103Google Scholar–25.

¹⁴ Ibid., 115.

¹⁵ Jervis, Perception and Misperception (fn. 1); Jervis, , “Deterrence and Perception,” International Security 7Google Scholar (Winter 1982–83), reprinted in Miller, Steven, ed., Strategy and Nuclear Deterrence (Princeton:Princeton University Press, 1984), 57–84Google Scholar; Jervis, “Introduction and Assumptions,” and “Perceiving and Coping with Threat” (both fn. 1).

¹⁶ Jervis, “Deterrence and Perception” (fn. 15), 58.

¹⁷ George and Smoke (fn. 1).

¹⁸ Lebow, Between Peace and War (fn. 1).

¹⁹ George and Smoke (fn, 1), 505.

²⁰ An excellent overview is Morgan, Patrick M., Deterrence: A Conceptual Analysis, 2d ed (Beverly Hills, CA:Sage, 1983Google Scholar).

²¹ Wohlstetter, Albert, “The Delicate Balance of Terror,” Foreign Affairs 37 (January 1959), 209CrossRefGoogle Scholar–34; Snyder, Glenn H., Deterrence and Defense (Princeton:Princeton University Press, 1961CrossRefGoogle Scholar); Ellsberg, Daniel, “The Crude Analysis of Strategic Choice,” The American Economic Review 51 (May 1961), 472Google Scholar–78.

²² For example, Russett, Bruce, “The Calculus ofDeterrence,” Journal of Conflict Resolution 7 (June 1963), 97–109CrossRefGoogle Scholar; Schelling (fn. 5), chap. 2.

²³ For example, Mearsheimer (fn. 1).

²⁴ More precisely, the initiator's subjective expected utility of attacking must be less than that of continuing the status quo. Among other factors, his expectations depend upon his subjective estimates of the expected costs of war, the probability of winning, and the estimated probability that the defender will retaliate. The latter, in turn, is a function of the initiator's subjective prior distribution over the defender's utilities in a game of incomplete information.

Analysts continue to struggle painfully for a fully satisfactory version of this game; as Harrison Wagner has remarked to us, “the rational theory of deterrence” doesn't exist. (Morgan [fn. 20, chap. 4], gives a clear exposition of the principal difficulty in the theory.) Hence the grim looks from game theorists when international relations scholars remark that there is no work left to do on rational deterrence. But just as one can believe Newton's laws without waiting for physicists to fully comprehend the nature of gravitation, so also the principal conclusions of a legitimate theory of deterrence are foreseeable even if the supporting arguments are at present incomplete; it is the former that we call “rational deterrence theory.”

²⁵ Schelling (fn. 5), 85–86.

²⁶ Jack Levy, “Quantitative Studies of Deterrence Success and Failure,” paper prepared for the meeting of the American Political Science Association, Chicago, IL, September, 1987, p. 2.

²⁷ George and Smoke (fn. 1), 72–77, 505.

²⁸ Ibid., 527.

²⁹ Many of these appeared first in the work of Schelling, Thomas, notably The Strategy of Conflict (Cambridge: Harvard University Press, 1960Google Scholar), and Arms and Influence (fn. 5).

³⁰ Kissinger, Henry, White House Years (Boston: Little, Brown, 1979), 208Google Scholar.

³¹ George and Smoke (fn. 1).

³² Jervis (fn. 1).

³³ Lebow, Between Peace and War (fn. 1), 304; see also “Conclusions” (fn. 4).

³⁴ Lebow, ' Between Peace and War (fn. 1).

³⁵ Ibid., 304.

³⁶ Lebow's later work is suggestive of auxiliary questions that arise from these considerations. He proposes the “general principle .. . that policy makers . . . are probably inordinately influenced by immediate and predictable costs” based on analysis of how the short-term interests of politicians diverge from the long-term interests of states. (Richard Ned Lebow, “Miscalculation in the South Atlantic: The Origins of the Falklands War,” in Jervis, Lebow, and Stein [fn. i], 123–24.)

³⁷ George and Smoke (fn. 1).

³⁸ George (fn. 6).

³⁹ George and Smoke (fn. 1), chap. 18.

⁴⁰ See, for example, Schelling (fn. 5), chap. 3.

⁴¹ The typological historical sequences in George and Smoke seem to us their most important contribution. However, the authors also give prominence to a list of more abstract propositions that derive from their work. Two typical examples:

Proposition 3: The initiator's belief that the risks of his actions are calculable and that the unacceptable risks of it can be controlled and avoided is, with very few exceptions, a necessary (though not sufficient) condition for a decision to challenge deterrence, i.e., a deterrence failure (George and Smoke, fn. 1, 529).

Proposition 6: Deterrence success will be favored but not ensured by the belief of the initiator that the defender possesses (a) an adequate and appropriate spectrum of capabilities; (b) sufficient motivation to employ them; and (c) probable freedom from impeding political constraints in the relevant time period (ibid., 530–31).

It is not clear to us that propositions of this kind are a challenge, or even an addition, to unitary rational-actor models of deterrence.

⁴² Ibid., 535.

⁴³ Ibid., chaps. 3, 4.

⁴⁴ Menger, , Investigations into the Methods of the Social Sciences with Special Reference to Economics (New York:New York University Press, 1985Google Scholar).

⁴⁵ Moe, Terry M., “On the Scientific Status of Rational Models,” American Journal of Political Science 23 (February 1979), 215CrossRefGoogle Scholar–43, for example, offers an interesting recent twist that might be profitably read for contrast with our position. He uses the traditional positivist covering-law model of explanation to attack a variety of bad arguments for rational choice models. The main issue, however, is whether rationality models use abstraction in the same way as any other scientific theory does. Moe is brief on this point, but he argues that natural science uses empirically based abstractions, while rational choice builds causal processes in by assumption. Unfortunately, he arrives at this conclusion by comparing rational choice theory, not to a natural science theory, but to an empirical generalization (Galileo's Law). Had he-more logically-compared the causal structure built into rational choice with the causal structure built into Newton's theory (which explains Galileo's Law), his conclusions would have been reversed.

Another approach to the role of formal theorizing, particularly in international relations, appears in de Mesquita, Bruce Bueno, “Toward a Scientific Understanding of International Conflict: A Personal View,” International Studies Quarterly 29 (June 1985), 121CrossRefGoogle Scholar–36.

⁴⁶ Jervis, “Introduction and Assumptions” (fn. 1), 5; Lebow, Between Peace and War (fn. 1), 102.

⁴⁷ Ibid., 19.

⁴⁸ Lebow (fn. 4), 232.

⁴⁹ George and Smoke (fn. i), 45.

⁵⁰ Ibid., 65.

⁵¹ When such theories appear, however, they are not likely to emerge directly from the results of case studies. In our view, the notion—common to both case-study and behavioral traditions-that theory emerges from masses of facts and lower-level generalizations, is false both to the history of the natural and social sciences and to the concept of theory.

⁵² George (fn. 6).

⁵³ George (ibid.), notes that Russett, Bruce, “International Behavior Research: Case Studies and Cumulation,” in Haas, Michael and Kariel, Howard, eds., Approaches to the Study of Polit ical Science (San Francisco:Chandler Publishing, 1970Google Scholar), makes a parallel case from a quanti tative perspective in favor of case studies for tracing causal patterns and for policy purposes. More recently, Ragin, Charles, The Comparative Method (Berkeley:University of California Press, 1987Google Scholar) has made similar claims. He argues that, at least as used by most social scientists, purely statistical methods rarely turn up intriguing findings or hypotheses. Social forces are nonlinear (“contextual”) and interactive (“holistic”). Hence they are blurred and averaged meaninglessly in “variable-oriented” approaches (or at least in the linear statistical models and casual data analysis in common use). He recommends case studies to find the detailed causal patterns in different contexts.

⁵⁴ Jervis, “Perceiving and Coping with Threat” (fn. 1), 13.

⁵⁵ George and Smoke (fn. 1), 516–17.

⁵⁶ Lebow, Between Peace and War (fn. 1), 11; see also Snyder and Diesing (fn. 1), 6.

⁵⁷ An excellent example of the difficulties of understanding the implications of different selection criteria is provided by the case of the 1958 Quemoy crisis, which Russett (fn. 23) treats as a deterrence success, George and Smoke (fn. 1) treat as a deterrence failure, and Le-bow, Between Peace and War (fn. 1), 13, excludes entirely because he judges it insufficiently “acute.”

⁵⁸ Huth, Paul and Russett, Bruce, “What Makes Deterrence Work? Cases from 1900 to 1980,” World Politics 36 (July 1984), 496–526CrossRefGoogle Scholar. Levy (fn. 26) has provided a broad discussion of this issue for the quantitative deterrence literature.

⁵⁹ Mearsheimer (fn. 1).

⁶⁰ Huntington, Samuel P., “Conventional Deterrence and Conventional Balance in Europe,” International Security 8 (Winter 1984), 38Google Scholar.

⁶¹ Heckman, James, “Sample Selection Bias as a Specification Error,” Econometrica 47 (January 1979), 153CrossRefGoogle Scholar–61; Maddala, G. S., Limited-Dependent and Qualitative Variables in Econometrics (Cambridge:Cambridge University Press, 1983CrossRefGoogle Scholar); Achen, Christopher H., The Statistical Analysis of Quasi-Experiments (Berkeley:University of California Press, 1986Google Scholar).

⁶² Starr, Harvey and Most, Benjamin A., “A Return Journey: Richardson, ‘Frontiers’ and Wars in the 1946–1965 Era,” Journal of Conflict Resolution 22 (September 1978), 441CrossRefGoogle Scholar–67.

⁶³ Ferris, Wayne, The Power Capabilities of Nation-States (Lexington, MA:Lexington Books, 1973Google Scholar); Weede, Erich, “Extended Deterrence by Superpower Alliance,” Journal of Conflict Resolution 27 (June 1983), 231CrossRefGoogle Scholar–54.

⁶⁴ Study designs like those of Ferris and Weede lack textbook purity, but the clarity of their procedures makes it possible to assess the direction of remaining biases unambiguously. It is clear that both estimate the effect of rational deterrence conservatively. That is, the clutter of irrelevant, nonconflictual dyads in the samples underestimates the effectiveness of deterrence. For example, suppose, with Weede, that alliance with the same superpower is estimated to reduce the dyadic probability of war by 10%. Then suppose that half the nation-pairs in the sample are not really relevant—that is, they never had hostile intentions toward each other. Then the 10% difference derives only from the relevant half of the sample, making the true difference twice as large, namely 20%.

⁶⁵ Luce, R. Duncan and Raiffa, Howard, Games and Decisions (New York:John Wiley, 1957) 31–32.Google Scholar

⁶⁶ Rapoport, Anatol, Two-Person Game Theory: The Essential Ideas (Ann Arbor:University of Michigan Press, 1966), 34Google Scholar.

⁶⁷ We are ourselves divided over whether international decision makers often do carry out expected utility computations to a reasonable approximation, or whether rational choice theory will usually be hopelessly inept at describing their conscious mental processes. Our point is that either view is consistent with the theory so long as their behavior follows rationality axioms.

⁶⁸ Friedman, Milton, “The Methodology of Positive Economics,” in Essays in Positive Economics (Chicago:University of Chicago Press, 1953Google Scholar).

⁶⁹ Steinbruner (fn. 2), 110.

⁷⁰ Stein, “A View From Cairo” (fn. 1), 55.

⁷¹ Jervis, Perception and Misperception (fn. 1); Betts (fn. 1).

⁷² For some striking examples, see ibid., 115–16, 121–25.

⁷³ See, for example, Orme's, John critique of Lebow's historical research in “Deterrence Failures: A Second Look,” International Security ii (Spring 1987), 96–124CrossRefGoogle Scholar.

⁷⁴ The relative unreliability of historical interpretation helps to explain why the quantitative literature on deterrence so often concentrates on power ratios and alliance bonds as causes of deterrence failure while ignoring mental events. The econometric theory of errors in variables demonstrated long ago that, to avoid serious inferential blunders, noisy data like self-reports of international decision makers must be discounted well beyond what common sense would suggest. This point is much less obvious within the informal inferential norms of the case-study tradition.

This does not imply that decision makers' self-reports need be problematic in all contexts. For example, survey research reports of preferences for political candidates will not ordinarily be subject to strong distortions, and they may therefore be used in tests of rational-choice assumptions. In general, tests of rational-choice theory based on reported preferences rather than on observed actions are tests of the joint proposition that the theory is correct and the reports are accurate. Only when we can have confidence in the latter proposition does the evidence bear on the rational-choice theory under consideration. (We are indebted to Henry Brady for this latter point.)

⁷⁵ The attentive reader will note that “unrebutted” does not mean “supported.” In partic ular, to read our hypothetical reconstruction of Kennedy's thinking as an attempt to provide evidence/or rational deterrence theory is to miss our point. What we are trying to demonstrate is that all such arguments are inherently flimsy in the absence of other evidence, and that such evidence is usually missing in the case-study literature, just as it is in our reconstruction.

⁷⁶ For example, Mearsheimer (fn. 1); Posen (fn. 1); Walt (fn. 1).

⁷⁷ Snyder, Jack L., “Richness, Rigor, and Relevance in the Study of Soviet Foreign Policy,” International Security 9 (Winter 1984–85), 89–108CrossRefGoogle Scholar.

⁷⁸ George and McKeown (fn. 9), 54.

⁷⁹ Russett (fn. 53); George (fn. 6).