Four new 15-item versions of the Boston Naming Test (BNT), a 15-item version used by the Consortium To Establish a Registry for Alzheimer's Disease (CERAD), and three 30-item BNT versions were studied in 26 subjects with Alzheimer's disease (AD) and 26 nondemented, neurologically normal controls. The four 15-item versions were statistically equivalent. On each version, controls performed significantly better than AD subjects, and scores on each could be extrapolated to a complete 60-item BNT score. The CERAD version also differentiated between AD and control subjects, but it was not equivalent to our four versions and could not be as easily extrapolated to a 60-item score. Even and Odd 30-item BNT versions were confirmed to be equivalent, and we further validated a 30-item Empirical Version designed to maximally discriminate between AD and normal subjects. Equivalent 15- or 30-item versions of the BNT will be useful in repeated assessments requiring independent forms of a naming task, as well as in situations where administration of the complete BNT is not practical.