The self‐referential encoding task (SRET)—an implicit measure of self‐schema—has been used widely to probe cognitive biases associated with depression, including among adolescents. However, research testing the stability of behavioral and electrocortical effects is sparse. Therefore, the current study sought to evaluate the stability of behavioral markers and ERPs elicited from the SRET over time in healthy, female adolescents (n = 31). At baseline, participants were administered a diagnostic interview and a self‐report measure of depression severity. In addition, they completed the SRET while 128‐channel ERP data were recorded to examine early (P1) and late (late positive potential [LPP]) ERPs. Three months later, participants were readministered the depression self‐report measure and the SRET in conjunction with ERPs. Results revealed that healthy adolescents endorsed, recalled, and recognized more positive and fewer negative words at each assessment, and these effects were stable over time (rs = .44–.83). Similarly, they reported a faster reaction time when endorsing self‐relevant positive words, as opposed to negative words, at both the initial and follow‐up assessment (r = .82). Second, ERP responses, specifically potentiated P1 and late LPP positivity to positive versus negative words, were consistent over time (rs = .56–.83), and the internal reliability of ERPs were robust at each time point (rs = .52–.80). As a whole, these medium‐to‐large effects suggest that the SRET is a reliable behavioral and neural probe of self‐referential processing.