Background: Memory self-efficacy (MSE) is the belief about one's mastery of memory functioning. In healthy elderly, memory complaints are related to MSE rather than to objectively measured memory capacity. MSE has scarcely been studied in patients that suffered a stroke. The aim of this study was twofold: (1) to examine whether memory capacity and MSE can predict the presence of memory complaints in stroke patients, and (2) to study which variables are the best predictors of MSE. Methods: In a cross-sectional study, 136 stroke patients (>18 months after onset) were recruited from April 2008 to November 2009. MSE was measured using the Metamemory in Adulthood questionnaire. Depression, coping and personality were measured using validated questionnaires, and memory performance was measured using the Rivermead Behavioural Memory Test (RBMT) and the Auditory Verbal Learning Test (AVLT). Patients were divided into a 'complaints' and a 'no complaints' group. Results: A lower MSE score was an independent predictor of having memory complaints (adjusted odds ratio: 0.422; p = 0.000), adjusted for age and depression. The RBMT and AVLT scores did not predict the presence of memory complaints (p > 0.263). Presence of memory complaints and depression were the strongest predictors of MSE (B = -1.748, p = 0.000; B = -0.054, p = 0.000), followed by word fluency, not having a partner and side of stroke (B = 0.038, p = 0.012; B = -0.517, p = 0.082; B = -0.479, p = 0.088). Conclusions: Memory complaints are predicted by MSE rather than memory capacity. MSE memory training might be an effective training strategy for reducing memory complaints in selected chronic stroke patients. Copyright © 2011 S. Karger AG, Basel.