The present study investigated the extent of mood-congruent false and true memory recognition in depression. A group of 25 patients with depression and 28 healthy controls completed a variant of the Deese-Roediger McDermott task. Four lists were read to participants in sequence, followed by a recognition task. The words in each list were associated with a central but unmentioned theme word that was either depression-relevant (i.e., loneliness), delusion-relevant (betrayal), positive (holidays), or neutral (window). Whereas it was expected to replicate the conventional mood-congruent effect in depression (better recognition of depression-relevant items), the available literature did not allow strong predictions to be made on the extent of mood-congruent false recognition in depression. Results showed that depressed patients learned emotionally charged material equally well as healthy participants but forgot significantly more neutral material. A conventional mood-congruent memory bias was not found, but relative to healthy controls, patients with depression committed more false recognition errors for emotionally charged words, particularly for depression-relevant items. The results confirm that depressed patients are biased toward emotional material. Reasons for the absence of the expected mood-congruent memory bias are discussed. It is suggested that researchers as well as clinicians should pay more attention to mood-congruent false recollection, because it may undermine the validity of autobiographic reports in depressive patients and may represent a maintenance factor for the disorder. Depression and Anxiety 00:000–000, 2005. © 2005 Wiley-Liss, Inc.