Self-compassion is assumed to have a protective role in the etiology of emotional problems in adolescents. This assumption is primarily based on correlational data revealing negative correlations between the total score on the Self-Compassion Scale (SCS) and symptom measures of anxiety and depression. Recently, however, the SCS has been criticized because this scale not only consists of items measuring compassionate self-responding (i.e., self-kindness, common humanity, and mindfulness), but also includes ‘reversed’ items measuring uncompassionate self-responding (i.e., self-criticism, isolation, and overidentification), which would undermine the validity of the scale as an index of a protective construct. The present article used two methods to demonstrate that compassionate (positive) and uncompassionate (negative) self-responding have differential effects on emotional problems in youths. In the first part, a meta-analysis based on 16 relevant studies demonstrated a modest protective effect of positive self-compassion on anxiety/depression and a large (and significantly stronger) vulnerability effect of negative self-compassion on such emotional symptoms. In the second part, network analyses were conducted on three previously collected data sets and these analyses again showed that negative self-compassion is more closely connected to young people’s symptoms of anxiety and depression than positive self-compassion. It is argued that the observed differential effects should not be discarded as a subversive fallacy, but rather offer an opportunity for studying the role of self-compassion in adolescents’ emotional psychopathology in a more sophisticated way, taking into account both protection and vulnerability.