Childhood loneliness is characterised by children's perceived dissatisfaction with aspects of their social relationships. This 8-year prospective study investigates whether loneliness in childhood predicts depressive symptoms in adolescence, controlling for early childhood indicators of emotional problems and a sociometric measure of peer social preference. 296 children were tested in the infant years of primary school (T1 5 years of age), in the upper primary school (T2 9 years of age) and in secondary school (T3 13 years of age). At T1, children completed the loneliness assessment and sociometric interview. Their teachers completed externalisation and internalisation rating scales for each child. At T2, children completed a loneliness assessment, a measure of depressive symptoms, and the sociometric interview. At T3, children completed the depressive symptom assessment. An SEM analysis showed that depressive symptoms in early adolescence (age 13) were predicted by reports of depressive symptoms at age 8, which were themselves predicted by internalisation in the infant school (5 years). The interactive effect of loneliness at 5 and 9, indicative of prolonged loneliness in childhood, also predicted depressive symptoms at age 13. Parent and peer-related loneliness at age 5 and 9, peer acceptance variables, and duration of parent loneliness did not predict depression. Our results suggest that enduring peer-related loneliness during childhood constitutes an interpersonal stressor that predisposes children to adolescent depressive symptoms. Possible mediators are discussed.