This study was undertaken to assess the most sensitive combination of tests to detect peripersonal unilateral neglect (UN) after stroke. The present study is a secondary analysis of a previously reported multicentric study of 203 individuals with right hemisphere damage (RHD), mainly subacute stroke, 11 weeks postonset on average, and 307 healthy controls. A battery of seven tests, providing 19 age- and education-adjusted z-scores, were given: the bells test, line bisection, figure copying, clock drawing, overlapping figures test, and reading and writing. Statistical analyses used a logistic regression and a receiver operating characteristic (ROC) curve after adjustment on demographic variables. A combination of four z-scores based on the following three tests provided good discrimination of patients with RHD from matched healthy controls: the starting point and the difference between the number of omissions on left and right sides from the bells test, rightward deviation in bisection of long lines (20 cm), and left-sided omissions in a reading task. The area under the ROC curve was 0.865 (95% confidence interval = 0.83-0.901), with sensitivity = 0.68, specificity = 0.95, accuracy = 0.85, positive predictive value = 0.90, and negative predictive value = 0.82. The most sensitive and parsimonious combination of tests to detect UN after stroke relies on four scores from three simple tests (bells test, line bisection, and reading). Future study is warranted to assess its ability to account for the functional difficulties of UN in daily life in the patient's actual environment.< Réduire