Background and purpose The Oxford Cognitive Screen is a stroke-specific screen to evaluate attention, executive functions, memory, praxis, language, and numeric cognition. It was originally validated in England for acute stroke patients. In this study, we examined the psychometric properties of the Dutch OCS (OCS-NL). Methods A total of 193 (99 acute stroke unit, 94 rehabilitation unit) patients were included in our study. A subset of patients (n = 128) completed a retest with the parallel version of the OCS-NL. Results First, we did not find evidence for a difference in prevalence of impairment between patients in the acute stroke versus rehabilitation unit on all but one of the subtests. For praxis, we observed a 14% lower prevalence of impairment in the rehabilitation than the acute stroke unit. Second, the parallel-form reliability ranged from weak to excellent across subtests. Third, in stroke patients below age 60, the OCS-NL had a 92% sensitivity relative to the MoCA, while the MoCA had a 55% sensitivity relative to the OCS-NL. Last, although left-hemispheric stroke patients performed worse on almost all MoCA subdomains, they performed similarly to right-hemispheric stroke patients on non-language domains on the OCS-NL. Conclusions Our results suggest that the OCS-NL is a reliable cognitive screen that can be used in acute stroke and rehabilitation units. The OCS-NL may be more sensitive to detect cognitive impairment in young stroke patients and less likely to underestimate cognitive abilities in left-hemispheric stroke patients than the MoCA. ; This work was supported by research grants from the Flemish Fund for Scientifc Research (FWO) awarded to H. H. (1171717 N, 1171719 N) and C. R. G. (G072517N, G0H7718N). R. B. is a senior post-doctoral fellow and R. L. a senior clinical investigator of the FWO.