抄録
Two experiments were conducted to investigate searcher's behavioral strategies in a two-person, zero-sum game called a search evasion game. Two boxes with the detection probabilities being identical were used and represented by two series of keys. In the Exp. I, four plays in which the detection probability was0.2, 0.4, 0.6and0.8 respectively were performed by37female college students. The results showed that the mean value of the trial numbers was greater than that of the game except for the cases where the probability of0.4was given. Luce's β model which is used in a two-choice problem was also applied to examine the correspondence between β-value and escape probability. The detection probability0.6resulted in higher estimation and that of0.4, lower estimation. In the Exp. II, 50plays were performed in succession by27female college students under the detection probability of0.4. Conditional entropy was calculated by selecting the plays which took five trials. In order to get stationarity, the plays were selected from the31st play to the50th play. The results showed that two or three succesive preceding trials influenced the player's later strategy, but the effect on the player's following strategy by one preceding trial was recognized to be negligible. This concludes that the distribution of search efforts was optimum at the time when a play was over, whereas the distrubtion in process of the play was not optimum.
