We studied psychometric properties of the Digit Span test and its nonverbal counterpart, the Knox Cube test, with reference to their utility in the clinical assessment of the elderly. Subjects were 100 residents of residential homes and semi-independent housing projects, aged 68 to 94. An administration of Digit Span with three trials per sequence length provides high reliability and minimizes floor effects, allowing separate interpretation of the Forward and Backward conditions. Due to its low internal consistency, the Knox Cube test cannot reliably distinguish differences among normal subjects, but may still be useful to detect impairment. Using regression equations, the Forward Digit Span score can be corrected for education level and both Digit Backward and Knox Cubes can be compared with Digit Span Forward.