We tested the reliability of brightness comparison testing in predicting relative afferent pupillary defects (RAPDs) in 206 patients. For unselected patients, brightness comparison testing correctly predicted the RAPD in 62% of patients. For subgroups of patients with glaucoma and optic nerve lesions, correct predictions were made in 92% and 83% of patients, respectively. Sensitivity and specificity of brightness comparison testing with a 6-V test light were 85% and 49%, respectively. In subgroups of patients who are strongly suspected of having an RAPD but in whom pupillary testing is not possible, brightness comparison testing can reliably predict presence or absence of an RAPD.