Mood congruity effects in induced mood states and affective disorders are well established. Recent evidence suggests that a similar process occurs in chronic pain patients, although the extent to which the memory bias is a consequence of the affective or sensory state of the subject in this group is unknown. In this study selective memory for sensory and affective pain-related information was investigated in depressed and non-depressed chronic pain patients and depressed psychiatric patients. A recall test comprising sensory, affective and neutral adjectives matched for frequency was followed by a recognition task, where the words of the recall test were randomized with an equal number of new adjectives matched for word type and frequency. Comparison of the three patient groups with normal controls revealed specific recall biases directly related to pain and depression in both chronic pain groups and the controls. Contrary to expectations, the depressed psychiatric patients failed to show a recall bias for affective adjectives: the possibility of cognitive avoidance as an explanation for this is discussed. Signal detection analysis of the recognition results suggested that selective memory in chronic pain and depression may to some extent be accounted for by differences in 'true memory', the contribution of response bias remaining less clear.