A. PROBLEM Recent studies on causal attribution take the basic assumption that individuals tend to explain the reasons for their own behavior as well as others’ behavior. The causal attributions are based on two different psychological mechanisms, which depend upon the agent of the behavior. While individuals attribute the others’ behavior through logical information-processing, they are likely to bias the attributions of their own behavior in a way to enhance or defend their self-esteem. They are inclined to attribute success to their personal dispositions(ability, effort) and failure to external forces(task difficulty, luck). A branch of the previous studies was focused on exploring how the information such as consensus, consistency, or distinctiveness affects the attributions. But their explorations are largely based upon experimental treatments, neglecting the individual differences of the subjects. Their main conclusion was that the information is selectively accepted or rejected on the bases whether it is consistent or inconsistent to the subject’s motivation to maintain self-esteem. It is reasonable that the conclusion can be refined when the individual differences of the subjects are considered. Field Dependence (FD) may be one of the significant variables to show the overgeneralization of the previous studies. It has been well validated that the field-independents perceive stressful situations independent from the emotion they have. A plausible hypothesis is that field-independents are less biased in attributing their success and failure, showing that the previous general conclusion should be applied with a certain boundary condition of the individual differences. B. METHODS A 2×3 factorial design was employed with the variations of achievement on a task(Success and Failure) and consensus information (High, Low, and None). 180 ROK Air Force officer cadets were randomly assigned to each condition. FD was measured in a group condition by Hidden Figure Test(HFT). Halfsplit reliability was acceptable (rxx´=.80). The Success and Failure conditions were individually manipulated by giving Ss different feedback of their performances in a testlike task called “Social Intelligence Test (SIT)”. The SIT consists of 30 items, requiring Ss to select the best associative word among four choices for a word. The Ss received the feedback of 80% correct answers in the Success condition and the feedback of 30% correct answers in the Failure condition. The consensus informations were varied by giving different norms of SIT. The Ss in the High consensus condition and Low consensus condition were respectively evaluated their performance as similar and dissimilar with others’ performances. Finally, a questionaire was administered to measure the dependent variables of the study. The Ss were required to attribute their performance according. to each of the four possible causal factors Cability, effort, task difficulty, luck). Each rating was made on a 7-point-scale. C. RESULTS The data confirmed the Self-serving Bias Hypothesis of previous studies. Ability attribution for success is significantly higher than that for failure. The data concerning the other factors such as effort, task, and luck are consistent with the hypothesis, although they are statistically insignificant. Two significant interactions between FD and attribution observed as hypothesized in this study. The regression slopes of F D on ability factor were significantly disordinated. The correlation was negative in the SuccessLow consensus condition whereas the correlation was positive in the Success-None consensus condition. The comparison of the regression slopes in the Success-None consensus condition and the Success-High consensus condition also yielded a significant disordinal interaction. Other regressions did not support nor contradict with the interactional hypotheses of this study. But some unanticpated trends were found. Interestingly enough, the regression slopes of FD on the dependent variables between the Success-None consensus condition and the Failure-None consensus condition were significantly varied, when ability factor and effort factor of the dependent variables were seperated. This results perhaps requires that those factors should not be considered as congruent under the same category of internal attribution. For this matter, further explorations are needed in the future.