This study aimed to objectively evaluate the severity of impulsivity [behavior inhibitory control (BIC) impairment] among adolescents with depression. In particular, those involved in non-suicidal self-injury (NSSI) behaviors, compared with those engaged in suicidal behaviors and adolescents without any self-injury behavior, using event-related potentials (ERPs) and event-related spectral perturbation (ERSP) within the two-choice oddball paradigm. Participants with a current diagnosis of major depressive disorder (MDD) engaged in repetitive NSSI for five or more days in the past year (n = 53) or having a history of at least one prior complete suicidal behavior (n = 31) were recruited in the self-injury group. Those without self-injury behavior were recruited in the MDD group (n = 40). They completed self-report scales and a computer-based two-choice oddball paradigm during which a continuous electroencephalogram was recorded. The difference waves in P3d were derived from the deviant minus standard wave, and the target index was the difference between the two conditions. We focused on latency and amplitude, and time-frequency analyses were conducted in addition to the conventional index. Participants with self-injury, compared to those with depression but without self-injury, exhibited specific deficits in BIC impairment, showing a significantly larger amplitude. Specifically, the NSSI group showed the highest value in amplitude and theta power, and suicidal behavior showed a high value in amplitude but the lowest value in theta power. These results may potentially predict the onset of suicide following repetitive NSSI. These findings contribute to substantial progress in exploring neuro-electrophysiological evidence of self-injury behaviors. Furthermore, the difference between the NSSI and suicide groups might be the direction of prediction of suicidality.