Objective: We examined the utility of the Birmingham Cognitive Screen (BCoS) in discriminating cognitive profiles and recovery of function across stroke survivors. BCoS was designed for stroke-specific problems across 5 cognitive domains: (a) controlled and spatial attention, (b) language, (c) memory, (d) number processing, and (e) praxis. Method: On the basis of specific inclusion criteria, this cross-section observational study analyzed cognitive profiles of 657 subacute stroke patients, 331 of them reassessed at 9 months. Impairments on 32 measures were evaluated by comparison with 100 matched healthy controls. Measures of affect, apathy, and activities of daily living were also taken. Between-subjects group comparisons of mean performance scores and impairment rates and within-subject examination of impairment rates over time were conducted. Logistic regressions and general linear modeling were used for multivariate analysis of domain-level effects on outcomes. Results: Individuals with repeated stroke experienced significantly less cognitive recovery at 9 months than those with a first stroke despite similar initial level of cognitive performance. Individuals with left hemisphere lesions performed more poorly than those with right hemisphere lesions, but both groups showed similar extent of recovery at 9 months. BCoS also revealed lesion-side-specific deficits and common areas of persistent problems. Functional outcome at 9 months correlated with domain-level deficits in controlled attention, spatial attention, and praxis over and above initial dependency and concurrent levels of affect and apathy. Conclusion: The study demonstrates how BCoS can identify differential cognitive profiles across patient groups. This can potentially help predict outcomes and inform rehabilitation.