In phase 1 of the study subjects high and low in trait anxiety were exposed to a novel mood manipulation technique. This involved the presentation of either pleasant or unpleasant newspaper photographs. Subjects exposed to the positive and negative manipulations reported corresponding state anxiety decreases and increases respectively. In phase 2, subjects were required to recall specific personal memories to anxiety-related and happiness-related cue words. It was found that the high trait anxiety group were faster in response to anxiety-related as compared with happiness-related cue words. Retrieval latencies were not affected by type of cue word for the low trait group. Further analyses showed that the two groups differed in terms of their retrieval latencies rather than their retrieval strategies. Multiple regression analyses revealed that, whereas state anxiety is related to an enhanced retrieval of memories to anxiety-related cue words, depression is associated with a slowing of responses to happiness-related cue words. The content specificity hypothesis is offered as an explanation for this pattern of results.