Aims Cognitive models propose that negative cognitive biases in attention (AB) and interpretation (IB) contribute to the onset of depression. This is the first prospective study to test this hypothesis in a sample of youth with no mental disorder. Methods Participants were 61 youth aged 9–14 years with no mental disorder. At baseline (T1) we measured AB (passive-viewing task), IB (scrambled sentences task) and self-report depressive symptoms. Thirty months later (T2) we measured onset of mental disorder, depressive symptoms and life events (parent- and child-report). The sample included children of parents with ( n = 31) and without ( n = 30) parental depression. Results Symptoms of depression at T2 were predicted by IB ( ß = .35, p = .01) but not AB ( ß = .05, p = .72) at T1. This effect was strongest for children who experienced multiple negative life events (F2,48 = 6.0, p = .018, ΔR2 = .08). IB did not predict depressive symptoms at T2 over-and-above the effect of depressive symptoms at T1 ( ß = .21, p = .13). Discussion These findings suggest that IB (but not AB) plays an important role in the aetiology of depression. Modifying IB may have a preventive effect on youth depression, particularly for youth who experience negative life events. This prospective study provides important foundations for future experimental studies.