¹ These studies, in addition to many of the articles in the present volume, can be grouped roughly in terms of their originating paradigms. One such is Deutsch, Karl W., and others, Political Community and the North Atlantic Area: International Organization in the Light of Historical Experience (Princeton, N.J: Princeton University Press, 1957)Google Scholar. Deutsch, his associates, protégés, and critics have produced a number of works particularly relevant for our topic, including Russett, Bruce, Community and Contention: Britain and America in the Twentieth Century (Cambridge, Mass: M.I.T. Press, 1963)Google Scholar; Jacob, Philip and Toscano, James (ed.), The Integration of Political Communities (Philadelphia: J. B. Lippincott Co., 1964)Google Scholar; Puchala, Donald, “International Political Community Formation in Western Europe: Progress and Prospect” (Ph.D. dissertation, Yale University, 1966, being revised for publication)Google Scholar; Deutsch, Karl W., ”Integration and Arms Control in the European Political Environment: A Summary Report,” American Political Science Review, 06 1966 (Vol. 60, No. 2), pp. 354–365CrossRefGoogle Scholar, and the several volumes written with Richard Merritt, Lewis Edinger, and Roy Macridis that this article summarizes; Alker, Hayward and Puchala, Donald J., “Trends in Economic Partnership: The North Atlantic Area 1928–1963,” in Singer, J. David (ed.), Quantitative International Politics: Insights and Evidence (New York: Free Press, 1968)Google Scholar; Inglehart, Ronald, “An End to European Integration?,” American Political Science Review, 03 1967 (Vol. 61, No. 1), pp. 91–105CrossRefGoogle Scholar; Fisher, William E., “An Analysis of the Deutsch Sociocausal Paradigm of Political Integration,” International Organization, Spring 1969 (Vol. 23, No. 2), pp. 254–290CrossRefGoogle Scholar.

The other major group of measurement and model-conscious regional integration studies take as their intellectual fount, the neo-functional approach of Haas, Ernst B. in his The Uniting of Europe: Political, Social, and Economic Forces, 1950–1957 (Stanford, Calif: Stanford University Press, 1958)Google Scholar and Haas, Ernst B. and Schmitter, Philippe C., “Economics and Differential Patterns of Political Integration: Projections about Unity in Latin America,” International Organization, Autumn 1964 (Vol. 18, No. 4), pp. 705–737CrossRefGoogle Scholar. Important, in part derivative, measurement studies include Lindberg, Leon, “The European Community as a Political System: Notes toward the Construction of a Model,” in Journal of Common Market Studies, 06 1967 (Vol. 5, No. 4), pp. 344–387CrossRefGoogle Scholar; Reinton, Per Olav, “International Structures and International Integration,” Journal of Peace Research, 1967 (Vol. 4, No.4), pp. 334–365CrossRefGoogle Scholar; Nye, Joseph S., “Comparative Regional Integration: Concept and Measurement,” International Organization, Autumn 1968 (Vol. 22, No. 4), pp. 855–880CrossRefGoogle Scholar; Barrera, Mario and Haas, Ernst B., “The Operationalization of Some Variables Related to Regional Integration: A Research Note,” International Organization, Winter 1969 (Vol. 23, No. 1), pp. 150–160CrossRefGoogle Scholar; and Schmitter, Philippe C., “Further Notes on Operationalizing Some Variables Related to Regional Integration,” International Organization, Spring 1969 (Vol. 23, No. 2), pp. 327–336CrossRefGoogle Scholar.

A careful reading of these and other works by Deutsch, Haas, Lindberg, and Amitai Etzioni reprinted in International Political Communities: An Anthology (Garden City, N.J: Anchor Books, Doubleday & Co., 1966) convinces me that the Deutsch and Haas groups overlap in too many ways to justify basically different labels for them.

² Alker, Hayward R. Jr, “Statistics and Politics: The Need for Causal Measurement,” in Lipset, Seymour M. (ed.), Politics and the Social Sciences (New York: Oxford University Press, 1969), pp. 244–314Google Scholar; and “Multivariate Data Analysis: Alternatives and Priorities,” paper presented at the Australian UNESCO Seminar on Mathematics in the Social Sciences, May 24, 1968.

³ Karl Deutsch, and others. See pp. 72–80 in International Political Communities for the details of their procedure for evaluating the contribution of various political appeals and amalgamation strategies. Since actual classifications are not fully reported, the present reconstruction cannot be assumed precisely to be the one finally favored by Deutsch and his coauthors.

⁴ These findings are worth comparing with Haas's hypotheses or prescriptions in his Beyond the Nation-State: Functionalism and International Organization (Stanford, Calif: Stanford University Press, 1964), p. 115ffGoogle Scholar, as to the most efficient strategies usable by integration-prone supranational bureaucrats.

⁵ Fisher, , International Organization, Vol. 23, No. 2, p. 265ffGoogle Scholar; Nye, , International Organization, Vol. 22, No. 4, pp. 874–876Google Scholar.

⁶ Nye himself has suggested some interesting alternatives. Recall his nonlinear restatement of. key federalist, functionalist, and neo-functionalist single-equation, unidirectional, partly causal hypotheses in International Organization, Vol. 22, No. 4, p. 875:

In words, if either policy, attitudinal, or security integration (PI₂, PI₃, PI₄, respectively), is to occur, both bureaucratic and jurisdictional components of institutional integration (PIB, PIJ) have to be high (above thresholds K₁, K₂).

Thus institutional (bureaucratic and jurisdictional) integration dichotomously defined (PI₁) is a function of the level of policy integration (PI₂) and a high threshold parameter K³, e.g., PI₁ is one when PI₂ is as high as K₃, zero otherwise (here the sign function translates “+” and no sign into “i,” “-” into “0” regardless of the levels of PI³ and PI⁴.

Verbally, attitudinal integration (PI₃), defined now as a continuous variable, increases above its early level because policy integration and institutional integration at some previous point t' reached high and intermediate level threshold values, K₃ and K₄.

Taken together, equations (5c) and (6d) suggest multiequation, reciprocal, and lagged relations between policy and attitudinal integration. Moreover, Nye has further suggested that high PI₂, and intermediate PI₁ will possibly lead to high PI₄ and higher PI₁, further complicating variable interdependencies. Thus if we want to use policy integration PI₂ to explain or predict or index attitudinal integration at a later date, or vice versa, we cannot be sure of the validity of such measures until we have specified and estimated to our satisfaction valid versions of the multiple relations between them.

If such multiequation systems as this are “recursively” or “separably” structured, then highly suggestive quasi-experimental validation procedures for single equations like a discontinuous version of (5a,b, or c) or (6d), discussed in Campbell, Donald T., “Reforms as Experiments,” American Psychologist, 04 1969 (Vol. 24, No. 4), pp. 409–429CrossRefGoogle Scholar, are relevant. Al Pelowski is now studying EEC innovations using such pretest-posttest research designs. I doubt, however, that such procedures could practicably be applied to simultaneously interdependent systems like (8a,b,c) below.

⁷ See Levy, Marion, The Structure of Society (Princeton, N.J: Princeton University Press, 1952)Google Scholar and Haas, Ernst, Collective Security and the future International System (Denver, Col: Social Science Foundation and Graduate School of International Studies, 1967–1968), Vol. 5Google Scholar. These writings are clearly within the “functionalist” tradition of sociology, but even though Ernst Haas has suggested interesting links between sociological functionalism and integration-relevant neo-functionalism in his Beyond the Nation-State, part 1, I shall hereafter avoid the sociological phrase “functional analysis” for the sake of clarity.

⁸ Note that (9a) is also not a completely adequate rendition of the necessity idea which may require a modal logic for adequate representation. E.g., “it is necessary that (Vc &…)” might be stated with a modal predicate “N” as N(Vc(t') &…); that such an approach would be appropriate here is also unclear, so we stick to a more conventional procedure of “overinterpreting” the proposition calculus and “” and “” in causal terms. A brilliant but advanced discussion of the problems in causal attempts to interpret either ordinary propositional logic or various modal logics is found in Simon, Herbert and Rescher, Nicholas, “Cause and Counterfactual,” in Philosophy of Science, 12 1966 (Vol. 33, No. 4) pp. 323–332CrossRefGoogle Scholar. It shows how the paradoxical counterfactual meaning of “” in equation (9a), that false I(t) is consistent with any V_c, R, MP combination, can be avoided in terms of a multiequation specification of linear or nonlinear causal relations between the variables involved. A similar argument holds for equation (9b) as well.

⁹ Cf. Haas, and Schmitter, , International Organization, Vol. 18, No. 4, p. 712Google Scholar:

On the basis of previous work on the political implications of economic unions it was found that a high rate of previous transaction, a similarity in size and power, a high degree of pluralism, and marked elite complementarity were extremely favorable to the rapid politicization of economic relationships.

Presumably Deutsch, and others, would fall back on such a formulation when not sure of the necessity or sufficiency of a particular condition.

¹⁰ The most relevant paper on alternate validation conceptions and procedures that I am aware of is Hermann, Charles, “Validation Problems in Games and Simulation with Special Reference to Models of International Politics,” in Behavioral Science, 05 1967 (Vol. 12, No. 3), pp. 216–231CrossRefGoogle Scholar. My own thinking has been considerably influenced by D. T. Campbell and Clyde Coombs; it will be elaborated in Greenstein, Fred and Polsby, Nelson (ed.), Handbook of Political Science (Reading, Mass: Addison-Wesley, forthcoming), Vol. 1Google Scholar. See the chapter on “quantitative methods.”

¹¹ Donald Puchala's article in this volume is a significant contribution to the growing literature in this regard, which cumulatively surely represents a measurement breakthrough. Edward Azar at Michigan State University is currendy putting together a volume of relevant papers. Earlier relevant efforts include Klingberg's use of judges from around the world in 1937 to estimate warlikeness among pairs of great powers. These were then subjected to multidimensional scaling procedures that revealed configurations strikingly like those characterizing the antagonists of World War II: Klingberg, Frank, “Studies in Measurement of the Relations among Sovereign States,” in Rosenau, James (ed.), International Politics and Foreign Policy: A Reader in Research and Theory (1st ed; New York: Free Press of Glencoe, 1961), pp. 483–492Google Scholar. Also, applications by Holsti, Loomba, and North of Osgood's semantic differential to international data using the General Inquirer: Ole Holsti, with the collaboration of Loomba, Joanne and North, Robert, “Content Analysis,” in Lindzey, Gardner and Aronson, Elliot (ed.), The Handbook, of Social Psychology, Vol. 2Google Scholar: Research Methods (Reading, Mass: Addison-Wesley Publishing Co., 1968)Google Scholar; Charles McClelland's development of a set of categories for describing international transactions that are both plausible and susceptible to high intercoder reliabilities: McClelland, Charles, “Access to Berlin: The Quantity and Variety of Events, 1948–1963,” in Singer, J. David, pp. 159–187Google Scholar; and the generalized Q-sort procedures painstakingly tested by Moses, Lincoln E., and others, “Scaling Data on Inter-Nation Action,” in Science, 05 26, 1967 (Vol. 156, No. 3778), pp. 1054–1059CrossRefGoogle ScholarPubMed. The core of the Q-sort procedure is placing act descriptions in separate piles along a specified continuum with frequency approximating a normal (bell-shaped) curve. Corresponding quantitative standardized scores are then assigned.

¹² Barrera, and Haas, , International Organization, Vol. 23, No. 1, p. 153Google Scholar.

¹³ A review of some of these alternatives is given in Torgerson, William, Theory and Methods of Scaling (New York: John Wiley & Sons, 1962)Google Scholar. See also Coombs, Clyde, A Theory of Data (New York: John Wiley & Sons, 1964)Google Scholar, and Alker, in Lipset, part 3.

¹⁴ For a further discussion of the SSARI procedures, the coefficient of alienation, and Ø, see Lingoes, J. C., “Recent Computational Advances in Nonmetric Methodology for the Behavioral Sciences,” Ann Arbor, Mich., 1966Google Scholar (mimeographed) and Guttman, Louis, “A General Nonmetric Technique for Finding the Smallest Coordinate Space for a Configuration of Points,” in Psychometrika, 12 1968 (Vol. 33, No. 4), pp. 469–506CrossRefGoogle Scholar.

¹⁵ If one is serious about estimating equations like (16), where the dependent variable must be zero or one, transformation procedures exist that will ensure a corresponding range of predicted values; cf. the discussion of probit analysis and related techniques in Goldberger, Arthur S., Econometric Theory (New York: John Wiley & Sons, 1964)Google Scholar.

¹⁶ At this point I had originally intended to discuss integration measures based on transactional indicators, as in Alker and Puchala, in Singer. Puchala's excellent review of such approaches in the present volume makes my own remarks largely redundant. From the validity-conscious perspective of the present treatment, it is perhaps worth suggesting an as yet untried validation procedure. If it is assumed that large absolute or relative volumes of freely chosen, reciprocal transactions are rewarding, then some lagged index like (8b) could be used to predict to subsequent declines in violence or its expectation. I have in mind using Klingberg's remarkably accurate judgments around 1937 of the likely outbreak of war among a 100 or so nation pairs as an important criterion variable for at least a partial test of this assumption using trade data for the 1920's and the 1930's. See Klingberg, in Rosenau.

¹⁷ Simon, Herbert, “A Formal Theory of Interactions in Social Groups,” chapter 6 of Models of Man (New York: John Wiley & Sons, 1957)Google Scholar. (Simon formalizes a number of relations verbally described in Homans, George C., The Human Group [New York: Harcourt, Brace & World, 1950]Google Scholar.) Most of the points we make in this section derive from Simon's analysis in the context of small group behavior.

¹⁸ It is of interest to compare this equation with a somewhat similar formula used by Lindberg, Leon and Scheingold, Stuart in Europe's Would-Be Polity: Patterns of Change in the European Community (Englewood Cliffs, N.J: Prentice-Hall, 1970), p. 114Google Scholar:

¹⁹ See, for example, the excellent parallel treatment of deterministic and stochastic dynamic models in Christ, Carl C., Econometric Models and Methods (New York: John Wiley & Sons, 1966)Google Scholar.

²⁰ Blalock, Hubert M. Jr, Causal Inferences in Nonexperimental Research (Chapel Hill: University of North Carolina Press, 1964)Google Scholar.

²¹ See the respective articles in Kempthorne, Otto, Multivariate Procedures in Biology (Ames: State University Press of Iowa, 1954)Google Scholar. The equivalence of present procedure to the more familiar Simon Blalock procedure of deriving relations among correlations is briefly discussed at a number of points in my “Causal Inference and Political Analysis,” in Bernd, Joseph L. (ed.), Mathematical Applications in Political Science (Dallas, Tex: Southern Methodist University Press, 1966), Vol. 2, pp. 7–43Google Scholar.

²² The concept of overidentification is discussed in most econometric texts: in Blalock; in Alker, in Bernd; in Christ; and at length in Fisher, Franklin, The Identification Problem in Econometrics (New York: McGraw-Hill, 1962)Google Scholar. Briefly it refers to modes which constrain reality, whose coefficients can be uniquely estimated with less data than is actually available, where there are, roughly, more independent equations than there are unknowns requiring estimation.

²³ The last chapter of Christ is a revelation in this regard. One can test a priori hypotheses (and confidence intervals) concerning specific nonzero coefficient magnitudes; one can test and compare models as to the statistical properties (e.g., autocorrelation) of their error terms, their predictions of observable time series, their goodness of fit, etc.

Two more difficult validation problems that I shall not discuss further concern the problem of comparing predictions from different causal orders and different causal variables. Thus one has to be especially careful about the relative plausibility of a variety of causal models. Reliance on prior predictions as to animate magnitudesand signs of coefficients are useful procedures in this regard. But when variables in one model do not exist in another, comparability in terms other than gross goodness of fit and randomness of residual errors becomes difficult. In the case of equations (17c) and (17d), the problem is easier in that both the linear and the nonlinear relationships are readily comparable. But when totally different conceptualizations exist, unless variable-to-variable translations are possible, cumulative comparisons are difficult.

²⁴ Assuming positive coefficients and certain plausible directional relationships (such as for a given D, greater support will in the long run produce greater I) and saturation effects (as S increases, integration activities I increase at a slower rate) Simon's most novel derivation is the existence of a threshold level of demands below which group integrated activities will decrease and disappear even though demand levels move up again near to the threshold. This phenomenon corresponds rather closely to the verbal theories of both Deutsch, and others and Haas, and Schmitter, , International Organization, Vol. 18, No. 4Google Scholar.

²⁵ Stability considerations are clearly exposited for the classic Richardson model (actually only one of many he discussed) in Rapoport, Anatol, Fights, Games and Debates (Ann Arbor: University of Michigan Press, 1960)Google Scholar, part I.

²⁶ The most relevant introduction and bibliography for the present exercise are Abelson, Robert, “Simulation of Social Behavior,” in Lindzey, and Aronson, ; Coplin, William (ed.), Simulation in the Study of Politics (Chicago: Markham Press, 1968)Google Scholar; Hermann, , Behavioral Science, Vol. 12, No. 3Google Scholar; Guetzkow, Harold, “Some Correspondences between Simulations and ‘Realities’ in International Relations,” in Kaplan, Morton (ed.), New Approaches to International Relations (New York: St. Martin's Press, 1968)Google Scholar. My own thoughts on the metaphysics of process modeling in international relations/domestic politics are best elaborated in “Computer Simulations, Conceptual Frameworks and Coalition Behavior” in Groennings, Sven, Kelley, E. W., and Leiserson, Michael (ed.), The Study of Coalition Behavior (New York: Holt, Rinehart and Winston, forthcoming)Google Scholar.

²⁷ For a variety of intriguing statistical procedures for partially validating complex models see Naylor, Thomas, Burdick, Donald, and Sasser, W. Earl, “Computer Simulation Experiments with Economic Systems: The Problem of Experimental Design,” Journal of the American Statistical Association, 12 1967 (Vol. 6, No. 320), pp. 1315–1337CrossRefGoogle Scholar. My general proposal is that when only aggregate data time series is available, one can summarize simulation performance in terms of causal statistical models applied to their outputs for real or hypothetical input. These models are in turn testable vis-à-vis the poorer real-world data.

²⁸ See Axelrod, Robert, Conflict of Interest (Chicago: Markham Press, 1970) for measurement proceduresGoogle Scholar.

²⁹ Although equations (29a,b) are but the beginning of an attempt to conceptualize integration in a synthetic way, the idea behind them is not new. Compare the somewhat analogous evolutionary concept of biological fitness. Biologists see this as involving various components: survival, longevity, time to age of first reproduction, and fertility. For bisexual populations, let x be the probability that an individual born at time o is still alive at time x, m_x be the number of female offspring born to a female during age x. If α is the age of first reproduction and ω the age of last reproduction, then the net reproduction rate Ro is

Where T is the mean generation rate, we can describe the corresponding growth rate exponentially in terms of a fitness coefficient r for a specific ecology as:

A revealing interpretation of r is given by a formula taking into account all the components of fitness mentioned above:

Here r represents the attrition rate from hostile forces against which species growth can just reproduce one female child per mother.

In another context Noam Chomsky has similarly argued that linguistic behavior can better be understood in terms of its competences than its performance. See his Aspects of the Theory of Syntax (Cambridge, Mass: M.I.T. Press, 1969)Google Scholar.

³⁰ The impoverishment of education by the demands of methodism poses a threat not only to so-called normative or traditional theory, but to the scientific imagination as well. It threatens the meditative culture which nourishes all creativity … [and] is the source of die qualities crucial to theorizing: playfulness, concern, the juxtaposition of contraries, and astonishment at the variety and subtle interconnection of things.

Wolin, Sheldon, “Political Theory as a Vocation,” American Political Science Review, 12 1969 (Vol. 63, No. 4), p. 1073CrossRefGoogle Scholar.