The purpose of this paper is to describe experiments or theoretical studies which are interested in the depressive mood effect on the emotional information processing. More precisely, our goal is to determine which factors govern the emergence of a memory bias in depression. Numerous authors revealed a phenomenon called "mood congruence memory" by which congruent information is better memorized than non-congruent information. Hence, this phenomenon means that memory efficiency is biased by the congruence of the material to memorize and the emotional state. The corpus of research in this area is considerable and our purpose is not to describe it exhaustively but to indicate the different methodological approaches. The first part of this paper deals with the presentation of the studies written by the late 1980s. Mood-congruent bias seems to be a reliable phenomenon in depressed subjects especially in explicit memory tasks (ie tasks where subjects are consciously trying to retrieve information in carrying out the tasks) such as free recall or recognition. By the 1990s, several authors developed an alternative cognitive view of depression, using Graf and Mandler's distinction between integration (ie activation or priming) and elaboration. According to these authors, integration is demonstrated when a past experience facilitates performance on a task which does not require deliberate recollection of that experience. In contrast, elaboration is a strategic process, comprising the linking of a word to other material in memory to form new relationships. Elaboration can be assessed by an explicit memory task such as free or cued recall. Taken together, a part of the results confirmed the presence of an explicit but not an implicit memory bias in depression. However, as Roediger and McDermott pointed out, the interpretation of non-significant findings in implicit memory tends to be uncertain. Watkins et al. themselves advocated a hint of a mood-congruity effect in the implicit task. Obviously, their implicit memory results required corroboration of other implicit memory measures. Recently, several others recent studies, using the same kind of tasks, have found evidence of an implicit memory bias in depression. So, it is apparent that the above studies have yielded variable findings. Thus, there is evidence indicating that several different memory processes may contribute to implicit memory tasks performance. Given this discrepant evidence of implicit memory bias, few authors decided to investigate the issue further and used a primed lexical decision task with both sub- and suprathreshold priming as a measure of implicit memory. Indeed, unlike the word completion task, it permits the separate the contributions of automatic and strategic processes. If priming occurs due to subthreshold presentation, when subject's awareness of the primes is restricted, then this would indicate that the priming effect is automatic and independent of conscious, strategic processes. On the other hand, if priming occurs with suprathreshold presentation (ie when primes are within awareness), then the priming effect may involve both automatic and strategic processes. In this view, Bradley et al. are the first who used a primed lexical decision task with both sub and suprathreshold priming to investigate the memory bias in depression. Results from these three studies indicate that non-clinical depressed individuals showed a depression-congruent implicit memory bias in the subthreshold but not in the suprathreshold priming condition, while clinically depressed individuals showed such a bias in both priming conditions. The study of Colombel et al., using a non-clinical sample, confirmed these results suggesting that the lack of depression-congruent effect in suprathreshold priming for non-clinical subjects might be due to the use of strategic processes which counteract the negative bias in automatic priming found in the subthreshold condition.