As risk for both non-suicidal self-injury and depression increases at adolescence, it is critically important to clarify equivocal findings regarding the temporal ordering of these health concerns. Addressing the limitations of past research, which included the use of only two data waves and analyses that do not account for within-person variance, the aim in this study was to examine the temporal ordering of non-suicidal self-injury and depressive symptoms during adolescence. Two independent samples were examined. One sample (first used by Marshall et al., 2013) consisted of 799 Swedish adolescents (11 to 15 years, Mage = 13.18, SD = 0.74; 51% girls). The second sample was a later cohort of adolescents from the same area (N = 2760; aged 12 to 18 years, Mage = 13.75, SD = 0.74; 49% girls). Analyses were two-part cross-lagged panel models (CLPMs) and random-intercept cross-lagged panel models (RI-CLPMs) to account for the semi-continuous distribution of NSSI and to determine if traditional lagged models were adequate for estimating within-person temporal ties. The results suggest that traditional CLPMs had inadequate fits to data. RI-CLPMs showed significant within-person, transactional associations between depressive symptoms and NSSI. Specifically, NSSI at a mean age around 13 years was associated with increases in depressive symptoms one year later, which were in turn associated with increases in NSSI when the average age was about 15 years old. Subsequently, lags suggested the potential for maintenance of comorbid distress. These results suggest that efforts to reduce depression and NSSI during adolescence need to begin in early adolescence, with a focus on adaptive responses to distressing emotional experiences.