The theories of a coalition formation appeared to be divided into two-one is derived from the viewpoint of psychology, and the other from the game theory. In this thesis, all the members were permitted an unlimited bargaining behavior, and also permitted to have enough chance of the intervention by the third person in order to make up the situations which were allowed to predict from the game theory viewpoint.Subjects were second-year junior high school students. Each group was composed of three persons, and there were a total of 15 groups. Types of triad used in this experiment were Type II (3-2-2), V (4-3-2), and VIII (4-2-2) proposed by Caplow, and each group, given a simple situation, played six games of each type, or 18 games in all. Every member was assigned random weight.The main results of our experiments were as follows:1. In Type II and V, there happened coalitions between two weaker members most frequently, thus confirming the Caplow's predictions. That is, players behaved basing on perception of initial strength. But in comparison with Vinacke & Arkoff's, the result of this study was closer to the chance distribution than theirs (Table 2).2. In Type VIII, coalitions were formed among two weaker members most frequently (Table 2).3. Each member made the same number of offers in Type II and V, while in Type VIII, weaker initiated offers significantly more often. And the number of offers to ally did not differ significantly between stronger and weaker in Type II and V, but differed significantly in Type VIII. And the number of receipts of alliance differed significantly between stronger and weaker in Type II and V, but did not differ in Type VIII (Table 4).4. In respect to the reward division in the initial offers from stronger to weaker, and from weaker to stronger, no differences were found between stronger's and weaker's, but there were significant differences among the three Types' players significantly more often offered equal split in Type II, and parity split in Type VIII (Table 7, 8).5. Concerning the kinds of reward division in the final coalition formation, parity split was found less than in the previous studies. And on weakers coalitions in Type II and VIII, there was considerable non-equal split in which the member of the coalition who contributed the most resources received the smaller share of the pay-off (Table 5).6. Regarding the pairs which came to an concurrence in the initial offer, whether the concurrence was continued to the end or not was examined. In short, the more difference in resources the pair had, the less the concurrence was continued (Table 9).7. The amount of paper communication on the bargaining was significantly less in Type VIII (Table 10, 11).