BACKGROUND: Substantial evidence suggests that cognitive deficits might persist after remission of a major depressive episode. However, results are inconsistent relative to the importance, pattern, severity, and moderating factors of this impairment. We aimed to determine how cognitive function following a major depressive episode compares with normal function, to specify the pattern and severity of persistent cognitive dysfunctions, and to examine the potential moderator effect of ten prespecified clinical and demographic variables. METHODS: We did a systematic review and meta-analysis of the published research. We searched systematically MEDLINE, Embase, PsycARTICLES, PsycINFO, the Cochrane Library, and relevant reviews identified by our database search, for research published from Jan 1, 1972, up to Jan 31, 2018, for studies of patients with past depression. We included all independent studies of patients who were assessed while in remission from a major depressive episode with at least one cognitive test, with inclusion of a healthy control group assessed with either the same test(s) as the major depressive episode group or with a standardised test with published age-stratified normative data. The main outcome was the difference in cognitive performance between major depressive episode remitters and healthy controls. Effect sizes were calculated using random-effects models for cognitive outcomes classified into 18 standard domains. Moderators of between-study variability were assessed using mixed-effects subgroup analyses and meta-regressions. FINDINGS: Of 10 126 citations identified by our search, 75 cognitive variables from 252 eligible studies (11 882 major depressive episode remitters and 8533 healthy controls) were included in our meta-analysis. Significant deficits following major depressive episode remission were observed in 55 (73%) of the 75 cognitive variables. These deficits (in the domains of processing speed, visual selective attention, working memory, verbal learning, and executive functioning), were generally small (30 [40%] of the 75 variables) or medium (22 [29%]) in size, although three long-term memory variables showed large deficits: g=-0·81 [95% CI -1·01 to -0·61] for logical memory immediate recall, g=-0·88 [-1·19 to -0·57] for logical memory delayed recall, and g=-0·84 [-1·18 to -0·50] for Cambridge Neuropsychological Test Automated Battery pattern recognition latency. Auditory attention, general autobiographical memory, inhibition ability unconstrained by speed, and intellectual functioning unconstrained by speed were equivalent between major depressive episode remitters and matched controls. The number of previous depressive episodes explained heterogeneity in the majority of variables (z=-2·06 [p=0·039] to z=-4·26 [p<0·0001]). INTERPRETATION: Deficits in selective attention, working memory, and long-term memory persist in remission from a major depressive episode and worsen with repeated episodes. Depression treatments, including relapse prevention, need to target these cognitive functions to optimise prognosis. FUNDING: None.