The present research examined the role of metacognitive confidence in understanding to what extent people's valenced thoughts guide their performance in academic settings. First, students were asked to engage in positive or negative thinking about exams in their major area of study (Study 1) or about themselves (Studies 2 and 3). The valence of these primary cognitions was manipulated to be positive or negative. Furthermore, a metacognitive variable, the perceived validity of the primary cognitions, was measured or varied to be relatively high or low. Finally, performance was assessed using a knowledge test (Study 1), a geometric shapes task (Study 2) or a selection of questions from the Graduate Record Examination (Study 3). In accordance with self-validation theory, we predicted and found that metacognitive confidence (relative to doubt) increased the impact of primary cognitions on performance. When thoughts were positive, increased confidence in the primary cognitions improved performance. However, when thoughts were negative, the same confidence validated the negative primary cognitions and reduced performance. Thus, metacognitive confidence can lead to opposite findings on performance depending on whether it validates performance-relevant positive thoughts or negative thoughts. Variations in the perceived validity of thoughts mediated the obtained effects. Therefore, we conclude that understanding the process of thought validation can help in specifying why and when metacognitive confidence is likely to work or to backfire in producing the desired performance effects.