Abstract Depression is associated with difficulties removing irrelevant negative material from working memory. Dysfunctions in updating the contents of working memory would lead to difficulties processing new information, and thereby make a depressive episode more likely. Cognitive deficit seems to be present in individuals suffering from unipolar disorder in the remitted state（Hasselbalch, Knorr & Kessing, 2011）. However, difficulty expelling negative material from WM in remitted depression is not well understood. This study was designed to assess whether the remitted individuals with recurrent depression are associated with a specific deficit in updating the contents of working memory that results in increased interference from irrelevant negative material. A modified Sternberg task was used to asses a person’s ability to update the contents of working memory in 21 remitted depression individuals and 21 normal controls. For each trial, participants were required to memorize 2 lists of emotional pictures and subsequently to ignore 1 of the lists. The impact of irrelevant emotional material on the ability to update the contents of working memory was indexed by response latencies on a recognition task in which the participants decided whether or not a probe was a member of the relevant list. Emotion material consists of 208 positive pictures and 208 negative pictures from the international affective picture system. The authors compared not only response latencies to probes from the irrelevant list to response latencies to the relevant probes of the same valence, but also response latencies to probes from the irrelevant list to response latencies to novel probes of the same valence (intrusion effect). An ANOVA on response times with group as the between-subjects factor and picture type (relevant, irrelevant) and picture valence (positive, negative) as within-subjects factors was conducted. The result showed that the analysis yielded significant main effect of condition (F (1, 40) = 12.18, p < .01), valence (F (1, 40) =12.04, p < .01) and an interaction effect of group, condition, and valence separately (F (1, 40) =9.79, p < .01). Interaction effect was further examined by analyzing response times of different valence. Compared to control participants, remitted individuals showed longer latencies to probes from the irrelevant list than from relevant probes when presented with negative pictures(t（20） = 3.33，p﹤.01). In contrast, normal participants exhibited longer latencies to probes from the irrelevant list than from relevant probes when presented with positive pictures(t（20）＝3.48, p﹤.001). In addition, another ANOVA on response times with group as the between-subjects factor and picture type (irrelevant, new) and picture valence (positive, negative) as within-subjects factors was conducted. The result yielded a significant interaction effect of group, condition, and valence (F (1, 40) =11.83, p < .01). Further analysis showed that remitted participants had longer latencies to probes from the irrelevant list than from novel probes when presented with negative pictures（t（20）＝4.88，p﹤.001）and had significantly greater intrusion effects ( t (40)=3.13，p﹤.01) compared with normal participants. However, the remitted participants show no intrusion effects when presented with positive pictures. These results indicate that remitted depression is associated with difficulties removing irrelevant negative material from working memory.