Resting-state studies conducted with stroke patients are scarce. First objective was to explore whether patients with good cognitive recovery showed differences in resting-state functional patterns of brain activity when compared to patients with poor cognitive recovery. Second objective was to determine whether such patterns were correlated with cognitive performance. Third objective was to assess the existence of prognostic factors for cognitive recovery. Eighteen right-handed stroke patients and eighteen healthy controls were included in the study. Stroke patients were divided into two groups according to their cognitive improvement observed at three months after stroke. Probabilistic independent component analysis was used to identify resting-state brain activity patterns. The analysis identified six networks: frontal, fronto-temporal, default mode network, secondary visual, parietal, and basal ganglia. Stroke patients showed significant decrease in brain activity in parietal and basal ganglia networks and a widespread increase in brain activity in the remaining ones when compared with healthy controls. When analyzed separately, patients with poor cognitive recovery (n = 10) showed the same pattern as the whole stroke patient group, while patients with good cognitive recovery (n = 8) showed increased activity only in the default mode network and fronto-temporal network, and decreased activity in the basal ganglia. We observe negative correlations between basal ganglia network activity and performance in Semantic Fluency test and Part A of the Trail Making Test for patients with poor cognitive recovery. A reverse pattern was observed between frontal network activity and the abovementioned tests for the same group. Hum Brain Mapp 35:3819–3831, 2014. © 2014 The Authors. Human Brain Mapping published by Wiley Periodicals, Inc.