Three studies tested whether self-doubt stems more from the absence of a strong desired self or the presence of a strong undesired self. Across studies, participants completed individual difference measures and then imagined a desired, neutral, or undesired possible self and completed strength measures for the imagined possible self. As predicted, compared to low self-doubt participants, high self-doubt participants reported less confidence in imagined desired selves and were slower to respond to desired self-consistent terms; however, they did not differ on explicit (confidence) or implicit (response speed) strength measures for imagined neutral or undesired selves. Moreover, the weaker desired selves imagined by high self-doubt participants predicted lower performance (compared to low self-doubt participants) on a final achievement test. Finally, the interactive effect of self-doubt and imagery on performance held after controlling for self-esteem and self-concept clarity but, consistent with predictions, was mediated by strength of the possible self.