【正文】 rable or unfavorable new information will lead to a catastrophic change in attitude. Empirical support for the CTA can be found not only in Latane and Nowak (1994) but also in Harton (1998), Harton and Latane (1997), and Liu and Latane (1995). Furthermore, the theory has been shown to be consistent with the contentions of other psychological theories including cognitive dissonance theory (Festinger, 1957), information integration theory (Anderson, 1981), thought polarization (Tesser,1976), and the elaborationlikelihood model (Petty and Cacioppo, 1986). This study considers whether CTA is useful in explaining the behavior of subjects in this , this study examines the following research question: R1: Will individuals who provide investment remendations exhibit hystersis in their subsequent assessments of the probability that an asset is impaired? Given that a major decision, such as an asset impairment decision, would be reviewed by external auditors, the impact of justification (a form of accountability) on the asset impairment decision is also examined in this study. In an extensive review of the accountability literature, Lerner and Tetlock (1999) conclude that accountability attenuates bias on tasks “to the extent that (a) suboptimal performance resulted from a lack of selfcritical attention to the judgment process and (b) improvement required no special training in formal decision rules, only greater attention to the information provided” (263). Similarly, in a review of the accountability literature in auditing, Messier and Quilliam (1992) conclude that accountability tends to increase the auditor39。 overall assessments of the potential asset impairment based on the mixed evidence provided. This variable is consistent with idea in SFAS No. 144 that assets need only be tested for impairment if there is sufficient reason to believe that an asset impairment is possible. DISCUSSION AND CONCLUSIONS The results provide some evidence of the applicability of the Catastrophe Theory of Attitudes (CTA) in the accounting domain. The parison of the probability estimates of the involvement group subjects with those of the control group subjects (see Figure 4) reveals some characteristics predicted by the CTA. In the CTA model, the attitudes of the subjects who are not involved in an issue will change in a linear fashion with changes in the positivity of information (., the magnitude of the normal factor). As individuals bee more involved, the pattern changes and bees discontinuous, perhaps even catastrophic. In Figure 4, the means for both groups of subjects appear to be changing in relation to the degree of positivity of asset impairment evidence. However, the involved subjects were slower than the control group subjects to incorporate the negative news into their probability assessments. This is consistent with CTA’s hystersis prediction. While the behavior exhibited by the involved subjects did not reflect a truly catastrophic change (as the CTA would predict for highly involved subjects), it does give some indication that involved subjects were reacting differently to the asset impairment evidence than the control group subjects. One reason why the differences are not dramatic may be that the involvement manipulation was not strong enough. In the context of the CTA model, the involvement manipulation may have only moved subjects a little to the front of the model on the involvement dimension (., the splitting factor) and, thus, created only small differences in the way subjects responded. Small differences in responses would be consistent with the CTA model because the discontinuous change in the attitude only develops as the level of involvement bees high. The finding that accountability mitigates involvement is consistent with prior studies showing that accountability reduces judgment biases. In every year except Year 4, involved subjects who had to justify their probability assessments estimated higher probabilities that the entertainment plex assets were impaired than the involved subjects who did not justify their assessments. Thus, accountability appears to attenuate the effects of the prior involvement in the asset investment decision. This finding is important because the CTA predicts a primacy effect. Acco。 Kaplan, 1996). These manipulations generally require subjects in the involvement or personal responsibility condition to make an initial decision as well as a second decision which all subjects make. The manipulation in this study required subjects in the involvement condition to prepare a remendation to the capital budgeting mittee immediately after reading the case materials describing the proposed investment in the plex. Subjects in the no involvement condition were not asked to advise the mittee on the investment decision. The involvement manipulation in this study was intended to force subjects to make a mitment to the project prior to considering any information about asset impairment. The accountability variable consists of two levels: accountability and no accountability for their decisions. The variable was operationalized by having subjects in the accountability condition justify their probability estimates. This type of manipulation is mon in the accountability literature as noted by Arnold (1997). Subjects were randomly assigned to one of four experimental groups: (1) a control group which did not make an investment remendation nor provide justifications for the probability assessments, (2) an involvement group which make an investment remendation prior to assessing the probability that the asset was impaired but did not justify their probability assessments, (3) an accountability group which justified (in writing) their probability assessments but did not make an investment remendation, and(4) a