1991 年 61 巻 6 号 p. 392-399
Previous studies using Bieri's ‘cognitive complexity’ score had supported ‘vigilance hypothesis’ which assumed that impressions of unfavorable persons were more complex than favorable persons. Thus, Bieri's measure seemed to be invalid because the findings were completely contrary to ‘frequency of interaction hypothesis’, presented in the theoretical framework of ‘cognitive complexity’, which assumed that impressions of familiar persons were more complex than unfamiliar persons, since people could be supposed to have more intimate acquaintance with favorable persons. The purpose of present study was to indicate the invalidity was caused by the research design where familiarity was dependent on favorability and to show that even Bieri's score could support ‘interaction hypothesis’, if one variables could be statistically orthogonized to the other. In each of two surveys reported, about two hundred female undergraduates completed a Rep test where they rated favorable and unfavorable persons on the basis of some dimensions which included favorability and familiarity. The results obtained through various regression analyses supported the above predictions. Moreover, it was revealed that, with favorability controlled, Bieri's score showed almost linear increases as familiarity increased, though the score of extremely unfamiliar persons was rather higher than the score the linear function could predict.