Objective: Due to a lack of time-efficient standardized assessments, there is a high risk of unidentified visual perception difficulties in stroke survivors. The Oxford Visual Perception Screen (OxVPS) is a 15-min performance-based screen for visual perception difficulties through tasks like picture naming and face recognition. This study evaluates the inter-rater reliability, convergent, and discriminant validity of OxVPS. Method: In this cross-sectional study, 161 stroke survivors within 8 weeks of their stroke, sufficient understanding of English, ability to concentrate for 15 min, and capacity to consent took part across three UK rehabilitation units. Video-recordings of OxVPS assessments were rated by an independent rater for inter-rater reliability. Convergent validity was assessed by comparing OxVPS scores with the Rivermead Perceptual Assessment Battery (RPAB), a 45-90-min battery of visual perceptual tasks. Discriminant validity compared OxVPS scores with performance on the Blind Montreal Cognitive Assessment (MOCA-B) for cognition and with the Visual Impairment Screening Assessment (VISA) for sensory vision. Results: Inter-rater reliability showed equivalent ratings (N = 107, t(106) = -14.77, p < .001) and mean difference of -0.01 point on a 10-point scale in a Bland-Altman analysis (95% confidence interval [CI]: -0.14 to 0.13). Convergent and discriminant validity demonstrated a high correlation of 0.78 (N = 58, 95% CI: 0.65-0.86) between OxVPS and RPAB, lower correlations of 0.52 with MOCA-B scores (N = 113, 95% CI: 0.37-0.64) and .39 with VISA scores (N = 110, 95% CI: 0.22-0.54). Conclusions: Data indicate good inter-rater reliability and evidence that OxVPS predominantly measures visual perception difficulties (convergent validity) in stroke survivors and less so cognition or sensory vision (discriminant validity).