¹ Waltz, Kenneth N., Theory of International Politics (Reading, Mass.: Addison-Wesley, 1979)Google Scholar; Singer, J. David, “Accounting for International War: The State of the Discipline,” Journal of Peace Research 18 (No. 1, 1981), 1–18.CrossRefGoogle Scholar

² Exceptions could readily be cited. Our contention is only that the practices outlined below are typical of much of what is done, and that they seem to reflect how scholars approach their research problems.

³ Analysts who work with the WEIS, CREON, and COPDAB events data sets adopt a different approach, of course. Rosenau's “pre-theory” of foreign policy and its various extensions and tests also quite clearly depart from the pattern by hypothesizing that different explanations may hold in different types of states. See Rosenau, James N., “Pre-Theories and Theories of Foreign Policy,” in Rosenau, , ed., The Scientific Study of Foreign Policy (New York: Free Press, 1971), 95–149Google Scholar; Rosenau, James N. and Hoggard, Gary, “Foreign Policy Behavior in Dyadic Relationships: Testing a Pre-Theoretical Extension,” in Rosenau, , ed., Comparing Foreign Policies (New York: John Wiley & Sons, 1974), 117–49Google Scholar; East, Maurice A., “Size and Foreign Policy Behavior,” World Politics 25 (July 1973), 556–76.CrossRefGoogle Scholar Our contention is only that such researchers tend to be the exception rather than the rule. It should also be said that a number of analysts report nongeneral findings; e.g., results that hold in the 19th century but not in the 20th, or that apply to major powers but not to minor ones. The point to note, however, is that the majority of these analysts initially look for—and apparently expect to find—general relationships. They have no initial theoretical expectation that their model should be applicable to some limited domain or that it should be useful only under certain conditions.

⁴ For research that adopts a narrow focus on war, see Most, Benjamin A. and Starr, Harvey, “Diffusion, Reinforcement, Geopolitics, and the Spread of War,” American Political Science Review 74 (December 1980), 932–46CrossRefGoogle Scholar, and Starr, Harvey and Most, Benjamin A., “Contagion and Border Effects on Contemporary African Conflict,” Comparative Political Studies 16 (April 1983), 92–117.CrossRefGoogle Scholar For a study dealing exclusively with alliances, see Job, Brian L., “Membership in Inter-Nation Alliances, 1815–1965Google Scholar: An Exploration Utilizing Mathematical Models,” in Zinnes, Dina A. and Gillespie, John V., eds., Mathematical Models in International Relations (New York: Praeger, 1976), 74–109.Google Scholar For a study focusing exclusively on arms races, see Philip A. Schrodt, “Richardson's Model as a Markov Process,” ibid., 156–75. Research dealing exclusively with international arms transfers is reported by Peleg, Ilan, “Military Production in Third World Countries,” in McGowan, Pat and Kegley, Charles W., eds., Threats, Weapons and Foreign Policy (Beverly Hills, Calif.: Sage Publications, 1980), 209–30.Google Scholar For an analysis focused exclusively on international negotiations, see Snyder, Glenn H. and Diesing, Paul, Conflict Among Nations (Princeton: Princeton University Press, 1977).Google Scholar Analysts who have written about “islands of theory” in international relations include Guetzkow, Harold, “Long Range Research in International Relations,” American Perspective 4 (Fall 1950), 421–40Google Scholar, and Forward, Nigel, The Field of Nations (Boston: Little, Brown, 1971).CrossRefGoogle Scholar

⁵ See, for example, East, Maurice A., “Status Discrepancy and Violence in the International System: An Empirical Analysis,” in Rosenau, James N., Davis, Vincent, and East, Maurice A., eds., The Analysis of International Politics (New York: Free Press, 1972), 299–319Google Scholar, and Wallace, Michael D., War and Rank Among Nations (Lexington, Mass.: Lexington Books, 1973).Google Scholar

⁶ See Coleman, James S., Introduction to Mathematical Sociology (Glencoe, Ill.: The Free Press of Glencoe, 1964), 516.Google Scholar

⁷ We are referring here to standard uses of the so-called “general covering law” or “refutationist” approach developed by scholars such as Hempel and Popper. For a brief rehearsal, see: Raymond, Gregory A., “Introduction: Comparative Analysis and Nomological Explanation,” in Kegley, Charles W. Jr, Raymond, Gregory A., Rood, Robert M., and Skinner, Richard A., eds., International Events and the Comparative Analysis of Foreign Policy (Columbia: University of South Carolina Press, 1975), 41–51.Google Scholar The reasoning underlying this approach can be crudely reconstructed in syllogistic form: If a “true” (i.e., universal) law exists between some concepts X and Y, then that law should always hold in the empirical world and should be evidenced by whatever are the appropriate associations between the respective operational indicators of the concepts, xi and yi. Empirical analyses are conducted to test for such associations between the occurrences of the xi and yi in the expectation that those results will inform us about whether or not a law actually exists between X and Y. The absence of the association between the xi and yi allows us, if the reasoning is sound, to reject the X/Y law, while the appearance of the anticipated xi/yi relationship allows us to say that the data are consistent with, or support the existence of, the X/Y law.

⁸ If y1, y2, …, yn are all potential means for attaining an assumed foreign policy goal, then each yi is a potentially valid indicator of Y (e.g., the initiation of an overt military strike or an increase in defense spending are complementary indicators of the concept “efforts to maintain or enhance security” or “foreign policy means”). Once this is recognized, the problem is easily understood: A focus on any one yi fails to provide full coverage of the range of possible behaviors in the “means” concept. Analysts who focus on only one yi are thus focusing only on “successful” occurrences. To that extent, an examination of the possible sufficiency relationship between any presumed causal factor (xi) and the yi is precluded. The point here is analogous to our criticism of analysts who focus only on actual occurrences of war. See Most, Benjamin A. and Starr, Harvey, “Case Selection, Conceptualization and Basic Logic in the Study of War,” American Journal of Political Science 26 (November 1982), 834–56CrossRefGoogle Scholar, and “Conceptualizing War: Consequences for Theory and Research,” Journal of Conflict Resolution 27 (March 1983), 137–59.

⁹ The reader may think that researchers could readily resolve this problem by constructing some composite scale of the various yi. In the past, the fashion would have been to factor analyze the yi. Such a strategy would work only if it is reasonable to assume that a given phenomenon yi invariably indicates the concept Y. As the discussion in the next section suggests, such a postulate might not be plausible. While the initiation of conflict may sometimes reflect the desire of decision makers to establish (or reestablish) their state's external security or viability, for example, the same behavior could reflect their desire to increase their government's viability in the face of domestic pressures.

¹⁰ Zinnes, Dina A., “Three Puzzles in Search of a Researcher,” International Studies Quarterly 24 (September 1980), 315–42CrossRefGoogle Scholar, at 323 (emphasis in original).

¹¹ Ward, Michael Don, “Research Gaps in Alliance Dynamics: Literature Review,” mimeo. (Evanston, Ill.: Northwestern University, 1980)Google Scholar, and Job, Brian L., “Grins Without Cats: In Pursuit of Knowledge of International Alliances,” in Hopmann, P. Terrence, Zinnes, Dina A., and Singer, J. David, eds., Cumuhtion in International Rehtions Research, Monograph Series in World Affairs, Vol. 18, Book 3 (Graduate School of International Studies, University of Denver, 1981).Google Scholar

¹² The need to reverse the tendency to attack international and foreign policy phenomena in fragmented, piecemeal fashion can be defended on other grounds. The point has been tellingly made by both Ostrom and Boynton. See Elinor Ostrom, “Beyond Positivism: An Introduction to This Volume,” in Ostrom, , ed., Strategies of Political inquiry (Beverly Hills, Calif.: Sage Publications, 1982), 11–28Google Scholar, and Boynton, G. Robert, “Linking Problem Definition and Research Activities: Using Formal Languages,” in Gillespie, Judith A. and Zinnes, Dina A., eds., Missing Elements in Political Inquiry (Beverly Hills, Calif.: Sage Publications, 1982), 43–60.Google Scholar Perhaps J. Bronowski, however, puts the point most clearly:

All science is the search for unity in hidden likenesses. … The scientist looks for order in appearances of nature by exploring such likenesses. For order does not display itself; if it can be said to be there at all, it is not there for the mere looking. There is no way of pointing a finger or a camera at it; order must be discovered and, in a deep sense, it must be created. What we see, as we see it, is mere disorder. … [Full grown] science grows from a comparison. It has seized a likeness between two unlike appearances. … The progress of science is the discovery at each step of a new order which gives unity to what seemed unlike (Bronowski, , Science and Human Values [New York: Harper Torchbooks, 1965], 13–15).Google Scholar

¹³ Harold, and Sprout, Margaret, An Ecological Paradigm for the Study of International Politics, Center of International Studies, Princeton University, Research Monograph No. 30 (Princeton, 1968)Google Scholar, and Harold, and Sprout, Margaret, “Environmental Factors in the Study of International Politics,” in Roseneau, James N., ed., International Politics and Foreign Policy, rev. ed. (New York: The Free Press, 1969), 41–56.Google Scholar

¹⁴ The careful reader will note that major elements of “Realism” and the “security dilemma” are being formalized in Table 2. See, for example, Morgenthau, Hans J., Politics Among Nations, 5th ed. (New York: Alfred A. Knopf, 1973)Google Scholar; Jervis, Robert, “Cooperation Under the Security Dilemma,” World Politics 30 (January 1978), 167–214CrossRefGoogle Scholar, and Keohane, Robert O., “Theory of World Politics: Structural Realism and Beyond” (Paper presented at the Annual Meetings of the American Political Science Association, Denver, 1982).Google Scholar These elements are being formalized around an expected utility framework that draws on the “cognitive behaviorism” of the Sprouts (fn. 13) in a manner eschewed by Realist theorists. While the model is somewhat similar to the Realist approach, therefore, it is also consistent with the Sprouts' dictum regarding environmental determinism: i.e., motivating factors need to be mapped through decision makers' perceptions even if all decision makers perceive alike and they do not make a (statistical) difference. Mapping through the decision makers is a logical requirement, not a statistical concern; nothing would happen (no relationships would hold) if decision makers did not exist. (We depart from Sprout and Sprout, however, insofar as we see policy initiations as “outcomes” of the interplay between what decision makers want to do and are capable of doing.)

^a Given the purpose of the illustration, it is not important to consider how one might operationalize either Cnt or Rnt. Only three points need to be noted. First, it should be understood that both capacities (C) and risks (R) are conceptualized here in terms of decision makers' perceptions. Put simply, a state's defense capacities and the risks it confronts are as they are recognized and understood by the decision makers. Consistent with this conceptualization, the capacities of “friendly” or distant states may be discounted by an nth state's decision makers in their calculation of Rnt; similarly, the capacities of friendly, and especially allied, states could be counted by the nth state's decision makers as supplemental sources of Cnt. Finally, it should be clear that capacities (C) and risks (R) may differ from the decision makers' objective capabilities or opportunities (see Postulate 1), which affect policy initiatives regardless of whether or not they are (accurately) perceived by the decision makers.

¹⁵ Another reason why states p and q are of interest derives from the so-called “empty cell” problem. Instead of ignoring the “nonbehaviors” of actors (the zeros or empty cells in data matrices), analysts might attempt to utilize the important information conveyed by such nonbehaviors when they construct and evaluate their models. For an empirical example of this phenomenon in which “neutral” war behavior plays the key role in model structure and evaluation, see Zinnes, Dina A. and others, “From War to War: A Stochastic Model of Alignment Behavior” (Paper presented at the Annual Meetings of the International Studies Association, Philadelphia, 1981).Google Scholar

¹⁶ Even though a universal or “true” relationship may exist, for example, difficulties with indicator validity and reliability could mask that association at the empirical level. In our own earlier work, we have suggested that additional complications may develop as a result of faulty case selection procedures or the failure to specify the correct logical linkages among the concepts and indicators (see fn. 8). Those problems are critical. They raise doubts about the standard approach (see fn. 7). To the extent that such difficulties exist, or at least cannot be ruled out, the results of the tests become ambiguous. The lack of empirical support in a given analysis could indeed imply that a supposed law is not true, but it could also mean that operation of a true law is obscured by the procedures adopted by the researcher conducting the test.

¹⁷ Bruce Bueno de Mesquita appears to have exactly this problem in mind when he considers research on the outbreak of war and observes, … the set of conditions sufficient for war may be so large that its specification is virtually impossible. … Most efforts to find the cause or causes of war focus on environmental circumstances that compel policymakers to wage war, but if we attempt to show causal relationships between environments and war, we are forced to ignore the role of national leaders and to act as though nations were no more than automatically reacting mechanisms (Mesquita, Bueno de, The War Trap [New Haven: Yale University Press, 1981], 4–5Google Scholar). It is this line of argument—combined with his contention that decision makers are not mechanistically compelled—that leads Bueno de Mesquita to conclude that the search for the sufficient conditions for war should be abandoned in favor of a search for necessary conditions. A similar position is suggested by Singer (fn. 1). Other related points are contributed by Kurth, James R., “A Widening Gyre: The Logic of American Weapons Procurement,” Public Policy 19 (Summer 1971), 373–404Google Scholar; and, Bertalanffy, Ludwig Von, General Systems Theory (New York: Braziller, 1968).Google Scholar

¹⁸ Expressed simply, the hypothesis is of the form, “Given A, if X (or any xi), then Y (or any yi”; i.e., given A, X is sufficient, but not necessary, for Y. If A were a measured variable rather than a theoretical postulate or axiom which is presumed to be either true or untrue, the proposition could be rewritten: “If A and X (or any xi), then Y (or any yi); i.e., A and X are jointly sufficient for Y.” Seen in this fashion, axioms are simply dichotomous variables that analysts either choose or are forced to leave unmeasured. This suggests that axioms should in general not be viewed as assertions of universal truth. They should instead be seen as antecedents in complex “if…, then …” statements which, in collapsed form, would read: “If the specified axioms are true for a given case, then the model should apply.” Once this is seen, we can move quickly beyond debates over the universal truth or falsity of the assumptions themselves: the question of the truth or falsity of the model is put aside. We can even move beyond concerns for generality and become more directly concerned with (a) identifying cases in which the axioms are true and in which the model should therefore be useful, and (b) developing complementary models that hold under alternative sets of initial axioms.

¹⁹ It is perhaps worth noting that this illustration entails an important simplifying assumption: states have the opportunity to act whenever they are willing to do so. Under the unified actor/national security dilemma formulation outlined in Table 2, we would expect either increasing external risks or decreasing defense capacity to contribute to the leaders' willingness to resolve the security problem if the change in either factor is sufficient to reverse previously positive Capacity-to-Risk inequalities. Whether or not the decision makers would act on that disposition, however, would be affected by their objective opportunities to act.

²⁰ See Boulding, Kenneth, Conflict and Defense (New York: Harper Torchbooks, 1962).Google Scholar

²¹ Technically, it is not necessary to use words and phrases such as “domain-specific” or “nice” in connection with laws. They are used here only to emphasize our point that analysts may err if they necessarily equate laws with universals (see fn. 22 below). The adjective “nice” is used in similar fashion by Boynton, G. Robert, “On Getting from Here to There: Reflections on Two Paradigms and Other Things,” in Ostrom (fn. 12), 29–68.Google Scholar

²² Because the points depart so sharply from accepted arguments and assumptions, it may be worthwhile to consider the following observations from scholars who use somewhat different terms, but come to identical conclusions:

There is a widespread misconception that theories are either “true” or “false.” A number of examples in physical science stand in direct contradiction to this. … Each of these theories or models is tautologically true, when their postulates are fulfilled. … They are not theories to be confirmed or disconfirmed in general, but only confirmed or disconfirmed in specific applications. As a result, they are not theories which explain “how people behave”; they are theories or models which describe how people behaved in this or that circumstance … one fruitful line of development … will be not to ask what is the theory of a certain kind of behavior, or what are the postulates which correctly describe a general area of behavior. The tactic proposed here is to set about developing and applying a number of sometimes-true theories which relate consequences to postulates, and which may adequately describe behavior in a given situation (Coleman [fn. 6], 516–18; emphasis in original).

… the laws of mathematics were invented too. The only demand made on them is that of consistency in a given context. But in different contexts, different laws may operate. Therefore, any statement of mathematics is valid or invalid not because certain relations are true or not true in the real world, but because these statements are or are not consequences of certain definitions and assumptions, which we are free to choose. The ambiguity of a mathematical statement results from a failure to specify with sufficient precision the exact context in which the statement is made (Rapoport, Anatol, Fights, Games and Debates [Ann Arbor: University of Michigan Press, 1960], 297–98Google Scholar; emphasis added).

Laws of nature … are different: to them the words “true,” “probable” and the like seem to have no application. … [We could] adopt Snell's formula tentatively, hypothetically, as a guide to further experiments to see whether the phenomena always happen so. On this level, we might ask “Is Snell's hypothesis true or false?” meaning, “Have any limitations been found to the application of his formula?” But very soon—indeed as soon as its fruitfulness has been established—the formula in our hypothesis comes to be treated as a law, i.e., as something of which we ask not “Is it true?” but “When does it hold?” When this happens, it becomes part of the framework of optical theory, and is treated as a standard. Departures from the law and limitations on its scope … come to be spoken of as anomalies and thought of as things in need of explanation (Toulmin, Stephen, The Philosophy of Science: An Introduction [New York: Harper Torchbooks, 1953], 78–79; emphasis in original).Google Scholar

Similar points have been made by Boynton (fns. 12 and 21), and Ostrom, (fn. 12) and “Introduction: Making Sense Out of a Muddle,” in Gillespie and Zinnes (fn. 12), 37–41.Google Scholar

²³ Rosenau, , ed., In Search of Global Patterns (New York: The Free Press, 1973)Google Scholar, Hopmann and others (fn. 11), and Zinnes (fn. 10). Scholars such as Singer and Russett have argued that analysts' collective research efforts are beginning to add up: Singer, J. David (fn. 1), and Singer, “Tribal Sins on the QIP Reservation,” in Rosenau (above), 167–73Google Scholar, and Russett, Bruce M., “International Interactions and International Processes: The State of the Discipline” (Paper presented at the Annual Meetings of the American Political Science Association, Denver, 1982).Google Scholar

²⁴ A similar point has been made recently in Bueno de Mesquita (fn. 17) and “Theories of International Conflict: An Analysis and an Appraisal,” in Gurr, Ted Robert, ed., Handbook of Political Conflict (New York: The Free Press, 1980), 361–98.Google Scholar

²⁵ Abel, , “The Element of Decision in the Pattern of War,” American Sociological Review 6 (December 1941), 853–59CrossRefGoogle Scholar; Sprout and Sprout (fn. 13).

²⁶ Russett, Bruce and Starr, Harvey, World Politics: The Menu For Choice (San Francisco: W. H. Freeman, 1981).Google Scholar

²⁷ Waltz (fn. 1), 37.

²⁸ Singer, himself a long-time proponent of the systemic level approach and the view that foreign policy decision makers may at least initially be presumed to be rather homogeneous, comes close to this point in reviewing the realpolitik, arms race, power transition, economic development, and imperialism models:

… at rock bottom the most important difference amongst the contending causes of war models is that of the foreign policy decision process. That is, each model assumes—often implicitly—a different class of decision makers in power and each postulates a different set of decision rules. … Note that we have assumed, to this juncture, a high degree of homogeneity in decision makers and the rules they employ, but to move closer to a full explanation, that assumption would have to be relaxed (Singer [fn. I], 14–15, emphasis added).

See also Singer, J. David, “The Level-of-Analysis Problem in International Relations,” in Rosenau (fn. 13), 20–29.Google Scholar

²⁹ Waltz (fn. 1), esp. pp. 37 and 65.

³⁰ Bueno de Mesquita (fn. 17).

³¹ One critique of Waltz (fn. 1) is that he has underestimated the cognitive links between social entities and their environments. The environment—the international system—is important. Types of international systems could (but need not) be conceptualized as operating to define domains such as the A and — A contexts dicussed in our example in the previous section. Nevertheless, Waltz artificially separates the “entity” and the “environment,” using the Sprouts' terms for the “ecological triad” (Sprout and Sprout, fn. 13). He seems to ignore the third leg of that triad, the entity-environment relationship, which must have a cognitive component. The “enduring anarchic character of international politics” (fn. 1, p. 66), which he regards as the central factor in the analysis, cannot explain the variance in behavior.