1 See Ball, Margaret, “Bloc Voting in the General Assembly,” International Organization 5 (1951), 3–31CrossRefGoogle Scholar, and Hovet, Thomas, Africa in the United Nations (Evanston, Ill.: Northwestern University Press, 1963Google Scholar).

2 Alker, Hayward R. Jr. and Russett, Bruce M., World Politics in the General Assembly (New Haven: Yale University Press, 1965Google Scholar).

3 Russett, Bruce, “Discovering Voting Groups in the United Nations,” American Political Science Review 60 (1966), 327–39.CrossRefGoogle Scholar

4 Newcombe, Hanna, Rose, Michael and Newcombe, Alan G., “United Nations Voting Patterns,” International Organization 24 (1970), 100–21.CrossRefGoogle Scholar

5 Powers, Richard, “United Nations Voting Alignments: A New Equilibrium,” Western Political Quarterly 33 (1980), 167–84.CrossRefGoogle Scholar

6 Martin-Bosch, Miguel, “How Nations Vote in the UN,” International Organization 41 (1987), 705–24CrossRefGoogle Scholar, and Iida, Keisuke, “Third World Solidarity,” International Organization 42 (1988), 275–95.CrossRefGoogle Scholar

7 Tolmin, Brian, “Measurement Validation: Lessons from the Use and Misuse of UN General Assembly Roll-call Votes,” International Organization 39 (1985), 189–206.Google Scholar

8 See McClelland, Charles D., “Let the User Beware,” International Studies Quarterly 27 (1983), 169–78.CrossRefGoogle Scholar This entire issue of International Studies Quarterly was given over to a critique of Events Data.

9 Schopen, Lynn et al., Nations on Record: United Nations General Assembly Roll-Call Votes (1946–1973) (Oakville, Ont.: Canadian Peace Research Institute, 1975Google Scholar).

10 Some readers may be concerned that using only 50 states will not be representative of a larger UN membership in 1975 and 1985. To check this I analyzed 1975 using 100 states (the programme maximum) and found very little change in the placement of the major powers or the locations or memberships of the blocs. This suggests that the votes of the 50 largest members capture the main dynamics of interaction in the General Assembly.

11 In SPSS-X, the MDS routine is called Alscal and the standard procedure uses a Euclidean distance model and a symmetrical proximities matrix. The two-dimensional solutions used here were iterated until the improvement in Young's S-stress statistic was less than .001. Aside from the two-dimensional representation, Alscal produced figures to show that the local minimum problem had been avoided.

12 Kruskal, Joseph and Wish, Myron, Multidimensional Scaling (London: Sage Publications, 1978), 45–48.CrossRefGoogle Scholar

13 This figure is shown in the raw solution form which emerges from the computer without interpretation. I have added circles to group nations on some of the other figures as discussed in the text.

14 See Newcombe et al., “United Nations Voting Patterns.”

15 This demonstrates the main problem with either R or Q factor analysis: both use factor scores from factors of different weights. For example, Alker and Russett's first factor, East/West, accounts for 53 per cent of the total variation while the second factor, North/South, only accounts for 13 per cent. Thus Figure 1 undoubtedly exaggerates distance in the vertical direction.

16 Jackson, Richard, The Non-Aligned, the UN and the Superpowers (New York: Praeger, 1983), 28.Google Scholar

17 To permit comparison with Powers’ analysis, which includes all votes, my 1975 data include all votes even when the majority was greater than 90 per cent. See Powers, “United Nations Voting Alignments.”

18 The particular SPSS routine is called “cluster analysis” but in fact is a hierarchical clustering which does not allow overlapping membership in clusters. Like MDS, this routine works from a difference measure (squared Euclidean distance) and iterates through stages. At the first stage it combines the two closest nations and calculates a new average distance between this cluster and the remaining nations. In each subsequent stage it combines two nations, a nation with an existing cluster or two clusters on the basis of their similarity. It continues this process until every item is combined in one cluster, although we usually stop considering the clusters at a point before the dissimilarities within the clusters are too large. See Lorr, Maurice, Cluster Analysis for Social Sciences (San Francisco: Jossey-Bass Publishers, 1983Google Scholar).

19 Kruskal and Wish, Multidimensional Scaling, 89.

20 Dimensionality has not been important to this analysis in that, unlike Alker and Russett, we are not interested in uncovering the underlying dimensions in the data. Nonetheless, an elbow test for dimensionality was conducted to explore the horseshoe phenomenon. In two to five dimensions no clear elbow was found in the declining stress level. This suggests that adding dimensions will not greatly improve the solution for these data and also confirms the horseshoe pattern of a strong first dimension.