¹ For an analysis of politics in terms of fights and games and for an application of game theory to the understanding of both types of conflicts see Rapoport, A., Fights, Games and Debates (An Arbor: University of Michigan Press, 1960)Google Scholar.

² By “democratic” I mean a system of government based on the right of all participants to form, leave and join groups of their choice for the purpose of selecting or deciding.

³ Mechanisms or situations other than equibalance can prevent rifts in society: for example, a supra-party consensus or overlaps within and between the political or other social systems. For an analysis of the role of overlap and consensus see Parsons, Talcott, “Voting and the Equilibrium of the American Political System” in Burdick, E. and Brodbeck, A. J., American Voting Behavior (Glencoe: Free Press, 1959)Google Scholar. While the tendency to equibalance seeks to obtain or return to a specific equilibrium, consensus and overlap, on the contrary, give room for disequilibriums at various points of the political system.

⁴ If the previous position is not remembered, random choices may produce a high level of apparent change. Our experiments were not affected by the failure of respondents to recollect previous positions, failure which often renders difficult the interpretation of answers to panel surveys of a cross section of the population. On this point see in particular Converse, Philip E. “The nature of belief systems in Mass Publics” in Apter, David E. (ed.), Ideology and Discontent (Glencoe: Free Press, 1964)Google Scholar; and Converse “Nouvelles dimensions de la signification des reponses dans les sondages sur les opinions politiques” Revue Internationale des Sciences Sociales, 16 (No. 1, 1964), pp. 21–38Google Scholar. The fact that all our successive votes took place within one hour and that the candidates had been given real names provided reasonable assurance that changers were actually aware that they were switching from one candidate to another.

⁵ In actual situations the “deviant” behavior of those who transfer support from one candidate to another would also have the effect of bringing the candidates closer to one another not only because of their obvious interest in maintaining as wide a support as possible but also because of the tendency of groups to resist trends toward the alienation of a member whose behavior is deviant. On this point see Dentler, R. A. and Erickson, K. T. “The Function of Deviance in Groups” Social Problems, 7 (Fall, 1959), pp. 98–107CrossRefGoogle Scholar.

⁶ See Duverger, M., Political Parties (London Methuen, 1965, ed.), pp. 215ff.Google Scholar

⁷ At least equibalance between three forces, since our experiments did not include a four-or-more candidate contest.

⁸ This tendency to the formation of groups of equal size differs from the movement of individuals between freely forming groups of the kind studied by James Coleman among pedestrians and playground children. See Coleman, James S. and James, John, “The Equilibrium Size Distribution of Freely Forming Groups” Sociometry, 24 (1961) pp. 36–45CrossRefGoogle Scholar. Because of the very rules governing our experiments the number of candidates could not grow. But the participants could have decided to abstain, which would have been one way of forming a group of one, akin to Coleman's solitary pedestrian. However in all our experiments the abstentions were so few (around 1%) that we ignored them altogether.

⁹ This tendency to change camp to rescue the underdog would seem to contradict one of the better established of small-group research findings: that individuals tend to conform to the norms of the group. See in particular Verba, S., Small Groups and Political Behavior (Princeton: Princeton University Press, 1961), p. 22Google Scholar; and Hopkins, T. K., The Exercise of Influence in Small Groups (Totowa1: Bedminster Press, 1964), pp. 118 ff.Google Scholar But, since no communication took place, the supporters of a given candidate were not known to one another, and voting for Smith or Jones could hardly be considered as a rite of passage into the group of the Jones or Smith supporters. When, however, the candidates were identified, by religion for example, and the un-balance was increased, the individual voter could be said to be under the social pressure of a group norm, such as “Protestants think that the Catholic underdog should not be rescued.” Even in that case it remains, however, that our groups, school children excepted, were not face-to-face groups. The tendency to restore equibalance should be expected to operate best in a situation of anonymity of the individual within social groupings lacking structure and leadership. These are the characteristics of the groups we studied, but also and increasingly so of party electorates in modern industrial societies.

¹⁰ Note also the difference between age groups in their first ballot choices. The younger children tend to vote for the candidate whose name appeared on top of the ballot. The grade eleven pupils divided their first ballot choices equally between the two candidates. No clear pattern emerged from the votes of the Canadian university students. The Seattle students showed a marked tendency to give their first ballot to the candidate listed second of three. Many indicated that their first choice was consciously made against the top candidate, simply because he was on top and thus expected to get a majority. On this point see the experiments reported by Schelling, showing that when a group of players are given some incentive to guess rightly what the majority of participants will individually decide, the culturally dominant side of an alternative is normally favored. For example: name “heads” or “tails”; if you and your partner name the same, you both win a prize. See Schelling, T. O., The Strategy of Conflict (Cambridge: Harvard University Press, 1963), p. 56Google Scholar.

¹¹ Whenever in our experiments an issue was injected to produce an expected identification between the voter and the candidate, such as sex or political party, the dominant transfers became those resulting from realignment behind the candidate with whom the voter had a common characteristic. Men behind the male candidate, Protestants behind the Protestant candidate, Democrats behind the Democratic candidate, etc. At this point in the experiment however differences appeared between sexes. In the U.B.C. psychology classes (Experiment 1), after the candidates had been identified by sex, only 9% of men transferred from the top to the less preferred candidate who had been identified as a woman, while 20% of women did so. Inversely 66% of men left the woman candidate, but a surprising 42% of women also did so. The political folklore finds here confirmation to its saying that women have, as candidates, a limited appeal to their own sex.

¹² For a study of the tendency to equibalance in the American party system see Sellers, C. “The Equilibrium Cycle in Two-Party Politics”, Public Opinion Quarterly, 24 (09, 1965), pp. 16–38CrossRefGoogle Scholar. For an analysis of the reasons, other than the tendency to rescue the underdog, which account for the breaking up of coalitions and alliances see Riker, William, A Theory of Political Coalitions (New Haven: Yale University Press, 1962)Google Scholar. Riker notes the danger inherent in the formation of too large a coalition: the disappearance of an opponent means the weakening of the bargaining power of those who, in the winning coalition, could have threatened to change camp; too large a coalition may also mean the dilution of rewards among the winners. For practical observations as well as theoretical formulations on coalitions see also Liska, George, International Equilibrium (Cambridge: Harvard University Press, 1957)CrossRefGoogle Scholar; and Kaplan, M. A, System and Process in Internation Politics (N.Y.: Riley, 1957)Google Scholar.