Over the last 2 decades, a substantial body of research has uncovered the impact bias, which reflect the tendency for people to overestimate the initial impact and/or duration of emotional event (Gilbert, Pinel, Wilson, Blumberg, & Wheatley, 1998；Wilson, Wheatley, Meyers, Gilbert, & Axsom, 2000；Levine, Lench, Kaplan, & Safer, 2012；Wirtz, Kruger, Scollon, & Diener, 2003) . Why do we persistently make these errors in our affective forecasts? One proposal is focalism, the other is unforeseen adaptation. However, the boundary conditions of focalism and unforeseen adaptation remain poorly understood. When is the focalism more important for the impact bias? To answer these questions, two experiments were conducted. With between-subject design in experiment 1 and within-subject design in experiment 2, participants were first asked to predict the affect after different events (being admitted to graduate in E1a, taking facture surgery in E1b, moving to new campus in E2), then experienced the events. All participants were randomly assigned to one of three conditions, that is, defocusing exercise, adaptation exercise, and control condition. The effects of three conditions on impact bias were compared. In experiment 1, the affect prediction and experience were compared by t - test, results showed that under the condition of defocusing exercise there is no significant difference between affect prediction and experience（E1a: t=0.94，p>0.05; E1b: t=1.67，p>0.05）, which means there was no impact bias. However, under the condition of adaptation exercise (E1a: t=2.24，p<0.05; E1b: t=2.82，p<0.01) and control condition（E1a: t=2.50, p<0.05; E1b: t=3.11，p<0.01）, the affect was overestimated, which means there was impact bias. In experiment 2, first in the participant level, the affect experience was regressed on affect prediction over six time point, in which the regression coefficient βreflected the accuracy of affect prediction. Then, the regression coefficients of three conditions (defocus exercise, adaption exercise, and control condition) were compared by ANOVA. Results showed that the defocusing exercise improved accuracy of affective forecast significantly, however, both adaptation exercise and control condition overestimate the effect of future events（F(2,151) =2.99，p<0.05）. Therefore, in the present study, the focalism is the main source of impact bias. Because the affective tasks of the present study provided clear situational clue, it concluded that when the situational clue is clear, focalism is the main source of impact bias. The present study has important theoretical contribution in that it clarified the boundary condition of focalism and unforeseen adaptation in the impact bias from the viewpoint of information using. The present study is also very important to reduce the impact bias, for example, when the situational clue is clear, the impact bias can be reduced significantly by defocusing exercise.