Based on hierarchical models of emotional disorders, relationships between higher- and lower-order components of anxiety and depression and emotion-congruent cognitive biases were examined. Two groups of participants (n = 189) were selected based on their scores on General Distress (the nonspecific factor of anxiety and depression). They performed an explicit memory test of incidentally-learned self-referenced material and an emotional Stroop interference task, using three types of stimuli: anxiety-related, depression-related and neutral non-valenced words. It was hypothesized that an attentional bias for anxiety-relevant words and a memory bias for depression-relevant words would be best predicted by anxiety-related and depression-related measures, respectively. Strengthening the notion that demonstration of these types of biases is not reliable in subclinical populations, both a correlational analysis as well as a more powerful extreme group analysis could not detect the existence of any emotion-related cognitive biases. (C) 2003 Elsevier Ltd. All rights reserved.