In my study, mentors matched with college mentees evaluated their self-efficacy nine times, during their participation in an academic mentoring program. Three distinct groups emerged as follows: (a) mentors who perceived themselves as moderately efficient throughout the mentoring relationship (the moderate stable (MS) group), (b) mentors who considered themselves moderately efficient at the beginning of the match, and increasingly so as the relationship progressed (the increasing (IN) group), and (c) mentors who perceived themselves as very efficient at the beginning of the match, but who subsequently experienced slight fluctuations of their self-efficacy (the high unstable (HU) group). Several personal and experiential factors such as the mentors' sensitivity to distress and the mentees' parental autonomy support predicted the likelihood of belonging to the IN or HU groups (as opposed to the MS group). These findings are interpreted according to the premises of the self-efficacy theory. Implications for mentor selection and training are discussed.