This study examined cognitive vulnerability in both depressed and non-depressed referred youngsters. Formerly depressed (FD) children and adolescents (n = 16) were compared to a currently depressed (CD) group (n = 18) on a self-referent encoding and memory task imbedded in a mood induction paradigm. In order to test the specificity of the findings to depression, the results of the FD were further compared with those of a clinical but never depressed (ND) group (n = 39) diagnosed with anxiety and/or disruptive behaviour disorders. The study confirmed the hypothesized differences between the groups in terms of self-referent encoding bias. Both the ND (p < 0.001) and FD (p < 0.001) group rated more positive words than negative words as self-descriptive while the CD endorsed a closer balance of positive and negative words (non-significant difference). No interaction effect was found for the recall task. The FD group evinced a similar memory bias than the CD group. However, also in the ND group, the number of proportional recalled positive words did not differ from the proportional recalled negative words. The findings yielded no evidence for a depression-specific in formation-processing bias. However, all subjects (FD, CD as well as ND) exhibited a memory bias and therefore 'clinical status' is considered as a cognitive vulnerability risk factor for developing a depressive disorder in the future. Copyright (c) 2008 John Wiley & Sons, Ltd.