¹ Spaeth, Harold J., “An Approach to the Study of Attitudinal Differences as an Aspect of Judicial Behavior,” Midwest Journal of Political Science, Vol. 5 (05, 1961), pp. 165–180CrossRefGoogle Scholar; Ulmer, S. Sidney, “Sealing Judicial Cases: A Methodological Note,” American Behavioral Scientist, Vol. 4 (04, 1961), pp. 31–34CrossRefGoogle Scholar, and also his “Homeostatic Tendencies in the United States Supreme Court,” in Ulmer, (ed.), Introductory Readings in Political Behavior (Chicago: Rand McNally, 1961), pp. 167–188Google Scholar.

² Fisher, Franklin M., “On the Existence and Linearity of Perfect Predictors in ‘Content Analysis’,” Modern Uses of Logic in Law, Vol. 1 (03, 1960), pp. 1–9Google Scholar; Kort, Fred, “The Quantitative Content Analysis of Judicial Opinions,” American Behavioral Scientist, Vol. 3 (03, 1960), pp. 11–14Google Scholar; Stuart Nagel, “Using Simple Calculations to Predict Judicial Decisions,” id., Vol. 4 (December, 1960), pp. 24–28; Tanenhaus, Joseph, “Supreme Court Attitudes toward Federal Administrative Agencies,” Journal of Politics, Vol. 22 (08, 1960), pp. 502–524CrossRefGoogle Scholar.

³ Schubert, , “Psychometric Research in Judicial Behavior,” Modern Uses of Logic in Law, Vol. 2 (03, 1962)Google Scholar, and “A Psychometric Model of the Supreme Court,” American Behavioral Scientist, Vol. 5 (11, 1961), pp. 14–18CrossRefGoogle Scholar.

⁴ Schubert, , “A Solution to the Indeterminate Factorial Resolution of Thurstone and Degan's Study of the Supreme Court,” Behavioral Science, Vol. 7 (Forthcoming, 1962)Google Scholar.

⁵ Coombs, Clyde H. and Kao, Richard C., “On a Connection between Factor Analysis and Multidimensional Unfolding,” Psychometrika, Vol. 25 (September, 1960), pp. 219–231CrossRefGoogle Scholar; Guilford, J. P., “Factorial Angles to Psychology,” Psychological Review, Vol. 68 (01, 1961), pp. 1–20CrossRefGoogle ScholarPubMed; and especially Coombs, A Theory of Data (forthcoming), chs. 1, 11, and 12.

⁶ Torgerson, Warren S., Theory and Methods of Scaling (New York: Wiley, 1958), pp. 352–359Google Scholar; Coombs, op cit., ch. 12.

⁷ See Kinsella v. Singleton, 361 U. S. 234; Grisham v. Hagan, 361 U. S. 278; McElroy v. Guagliardo, and Wilson v. Bohlender, 361 U. S. 281; all decided January 18, 1960. Cf. Schubert, , “Attitudes towards Civilian Control and Stare Decisis in the Warren Court,” in Schubert, (ed.), Judicial Decision-Making, Vol. IV in the series of International Yearbooks in Political Behavior Research (New York: The Free Press of Glencoe, Forthcoming, 1962)Google Scholar.

⁸ 338 U. S. 25 (1949).

⁹ Mapp v. Ohio, 367 U. S. 643 (June 19, 1961).

¹⁰ Concurring opinion in Irvine v. California, 347 U. S. 128, 138–139 (1954).

¹¹ See Schubert, , “Policy without Law: An Extension of the Certiorari Game,” Stanford Law Review, Vol. 14 [(03, 1962), pp. 284–327]CrossRefGoogle Scholar.

¹² Pritchett, C. Herman, The Roosevelt Court: A Study in Judicial Politics and Values, 1937–1947 (New York, 1948)Google Scholar; Schubert, , Quantitative Analysis of Judicial Behavior (Glencoe, 1959)Google Scholar.

¹³ The phi coefficient is an approximation of the Pearsonian r correlation coefficient, and is appropriate to use when, as here, the two distributions to be correlated reflect a genuine dichotomy. The phi coefficient is relatively simple to compute: it is the ratio of the difference of the cross-products of the diagonals of a four-fold table, to the square root of the product of the marginals. For the data of Table II,

Evidently, the sign of the coefficient depends upon which diagonal cross-product is the larger; or, in other words, upon whether or not a pair of justices agree more often than they disagree, and also upon whether they agree both in assent and in dissent, i.e., whether their disagreement is divided equally or disproportionately between the cells of the diagonals. The maximum range of the phi coefficient is from + 1 to − 1; but these limits rarely are attained empirically, since the maximum size of phi is a function of the distribution of the marginals, and can be ±1 only when all four marginal frequencies are equal. See Cureton, Edward E., “Note on ϕ/ϕ _max,” Psychometrika, Vol. 24 (03, 1959), pp. 89–91.CrossRefGoogle Scholar

¹⁴ When factor analysis is performed by hand use of a calculator instead of utilizing a computer program, it is necessary to arrange the correlation coefficients in the form of a square symmetrical matrix, with the major diagonal filled with the estimates of the highest communality for each justice. For the techniques of factor analysis the interested reader is referred to any of the several standard works on this subject, e.g., Fruchter, Benjamin, Introduction to Factor Analysis (Princeton: Van Nostrand, 1954)Google Scholar; Cattell, Raymond B., Factor Analysis: An Introduction and Manual for the Psychologist and Social Scientist (New York: Harper, 1952)Google Scholar; Thurstone, Louis L., Multiple-Factor Analysis (Chicago: University of Chicago Press, 1947)Google Scholar.

¹⁵ I am indebted to John P. Gilbert for developing a program for the computation of the phi coefficients, and also one for the computation of the factorial distances discussed in the final section of this paper. The computer program for the IBM 650 utilized was Ceniroid Factor Analysis, Ident, No. SU-4524, written by Jonathan E. Robbins (April, 1957), developed and tested by the Watson Scientific Computing Laboratory, Columbia University, and modified by Mr. Gilbert, Center for Advanced Study in the Behavioral Sciences, Stanford (multilithed).

¹⁶ E.g., Fred Kort in this Review, Vol. 54 (December, 1960), p. 1000.

¹⁷ Perhaps it should be noted, for the benefit of readers not familiar with the method, that orthogonal axes are statistically independent, while oblique axes are correlated with each other; therefore, making a factor interpretation based directly upon a system of orthogonal axes implies an assumption that there is no relationship among the factors, which must be conceived to be independent of each other. Applied to the present data, this would involve the assumption that there was no relationship, at least in the minds of the justices, among the major issues of public policy toward which they responded in their voting.

¹⁸ Schubert, op. cit. note 12 supra, pp. 270–290.

¹⁹ Civil Liberties and the Vinson Court (Chicago: University of Chicago Press, 1954), p. 227Google Scholar.

²⁰ Time, Vol. 77 (06 16, 1961), p. 18Google Scholar.

²¹ This emphasis is reflected, for example in Pritchett, ibid., p. viii; and in McCloskey, Robert, “Deeds Without Doctrines: Civil Rights in the 1960 Term,” in this issue of this Review, above, pp. 71–89Google Scholar. In his scries of annual survey articles on constitutional law, published in this Review during the decade 1952–61, David Fellman allocates three-fourths of the approximate 370 pages, in which he discussed civil liberties, economic, and taxation issues, to what are here denominated C scale cases, and the remaining one-fourth to E and F scale cases.

²² Torgerson, op. cit., p. 324.

²³ The use of this formula was kindly suggested by John P. Gilbert.

²⁴ The direction of orthogonal factor axes extracted by the complete centroid method is arbitrary; the asterisks following the identifying numbers for the first and second axes (i.e., I* and II*) signify that the polarity of these two axes was reversed to facilitate uniformity in inter-term comparisons in the larger study to which this article relates.

²⁵ Coombs, op. cit., ch. 12; and cf. Coombs and Kao, op. cit., p. 230.

²⁶ However, it will be observed that the probability of such perfect agreement between the two sets of rankings having occurred by chance alone is less than one chance in 40,000.

²⁷ In an article devoted to the attitudinal similarities and differences of these two justices, Spaeth has concluded that his data, also, “show significant ideational difference between Black and Douglas.” Op cit., p. 176.