Two studies examined the Warrington Recognition Memory Test (RMT) discrepancy index (Words-Faces) in a large sample of patients heterogeneous with respect to age, education, gender, and neurological diagnosis. In Study 1 (N = 504) we used cutoffs from the Words-Faces discrepancy scores derived from Warrington's original validation sample to attempt to accurately classify patients with left, right, or diffuse brain damage. Sensitivity for left hemisphere patients (Faces > Words) was 10% with a specificity of 88%, whereas sensitivity for right hemisphere patients (Words > Faces) was 48% with a specificity of 86%. For patients with diffuse brain damage (Words = Faces) sensitivity was 69% and specificity was 38%. In Study 2 (N = 263), we examined the relationship between the Words-Faces discrepancy score and Wechsler Memory Scale-Revised (WMS-R; Wechsler, 1981) Logical Memory and Visual Reproduction subtests. Contrary to predictions, patients with Words > Faces performed better on both WMS-R subtests; the Faces > Words discrepancy was not related to Visual Reproduction performance. Potential reasons for these negative findings are discussed, as well as cautions for future RMT discrepancy index use.