The present investigation examined the validity of the revised Hopkins Verbal Learning Test (HVLT-R). In a principal components analysis with varimax rotation, measures of new learning and delayed recall loaded on a single factor distinguishable from measures related to general cognitive function and visual memory. The HVLT-R also correlated most strongly with other tests of verbal memory and relatively weakly with a test of general intelligence. Group comparisons showed that normal controls performed better than age- and education-matched patients with probable Alzheimer's disease (AD) or vascular dementia (VaD). Discriminant function analyses and Bayesian statistics revealed high classification accuracies for dementia patients versus controls. When scores on the HVLT-R and other neuropsychological tests were subjected to discriminant function analyses, performance on the HVLT-R delayed recognition task was found to be the most useful in discriminating patients with AD from those with VaD. We conclude that the HVLT-R is a valid test of verbal learning and memory that is best suited for use with elderly patients suspected of dementia.