In the present study we examined the basis of students' self-evaluations of comprehension performance. We manipulated text difficulty to maximize performance variability across reading trials and to minimize the likelihood of giving similar performance evaluations across trials. Students were allowed to acquire task (prior) experience across nine reading trials. A path model analysis was then used to examine the basis of students' self-evaluations within and across trials 10, 11, and 12. The results showed that despite nine learning trials, students did not assess performance based on each individual trial (proximal experience). Rather, self-evaluations were influenced by prior trials (distal experience). Thus, immediate self-evaluations are not assessments of immediate performance but assessments of generalized self-representations of ability formed from distal experience.