Time is a framework of understanding and defining self. Individuals could mentally travel back to the past and forward to the future. Research has demonstrated that the self drives mental time travel, showing self-descriptions predict the valence and personal focus of future events, and self-concept (e.g. core self-evaluations, self-construal, self-efficacy ) can drive the construction of past and future events. However, the cognitive mechanism of the effect of mental time travel on the self wasn’t explored yet. Task self-confidence in the present study means the extent of individuals’ sureness about their performance in a task, and it’s a part of the foundation of the cognitive construction of self-confidence. The current study examined the effect of time direction and events’ emotional valence on task self-confidence. This study included 2 experiments. In Experiment 1, 40 undergraduates were randomly assigned to 2 groups (past and future). First, they should recall or imagine self-related events freely in 8 time periods respectively (e.g. the 8 time periods of the past group were last year, the year before last year, 3 years ago, 4 years ago, 5 years ago, period of middle school, period of primary school, preschool period), and described them in the blanks below each time period. Afterward, participants predicted their performance in the upcoming graphic reasoning test by rating 7 items. All the 7 items’ average score represented the task self-confidence. The procedure of Experiment 2 was similar to Experiment 1. Eighty undergraduates were randomly assigned to 4 groups (i.e. past-positive, past-negative, future-positive, future-negative). According to what was required to each group, participants should recall or imagine self-related positive or negative events freely in 4 time periods respectively (e.g. the 4 time periods of the 2 past groups were last year, 3 years ago, 5 years ago, period of primary school), and described them in the blanks. After describing each event, they rated their feelings about it from 1(extremely negative) to 7(extremely positive). To eliminate order effect, the orders of generating events were balanced. Then, participants completed the prediction task which was the same as that in Experiment 1. Results of Experiment 1 showed that: time direction affected task self-confidence indirectly through the mediating effect of events’ emotional valence, that was, compared with events participants recalled in the past group, events in the future group were more positive, so that they induced higher task self-confidence. And in the model consisting of time direction, events’ emotional valence and time direction × events’ emotional valence, only events’ emotional valence could predict task self-confidence significantly. If events that people generated were more positive, they would introduce higher task self-confidence. When events’ emotional valence was manipulated, the effect of time direction and events’ emotional valence was separated (Experiment 2), and results showed: time direction didn’t affect on task self-confidence. A 2 (time direction: past, future) × 2 (events’ emotional valence: positive, negative) ANOVA on task self-confidence showed a significant main effect for events’ emotional valence, and a significant time direction × events’ emotional valence interaction. When participants recalled past events, valence didn’t affect task self-confidence; while as to future events, valence positively affected task self-confidence, positive events induced higher task self-confidence than negative events. When participants recalled or imagined positive events, time direction didn’t affect task self-confidence; while as to negative events, time direction negatively affected task self-confidence, that was, future direction induced lower task self-confidence than past direction. These results indicate that both time direction and events’ emotional valence can influence task self-confidence.