This study evaluated cross-sectional and longitudinal age relationships, after age 60, for three subtests of the Wechsler Memory Scale: logical memory, associate learning, and visual reproduction. Cross-sectional regression analyses indicated that age, sex, race, and education accounted for approximately 20 to 35% of the variance in test scores, with education the strongest predictor. Longitudinally, consistent linear declines were found only for hard associates and visual reproduction. Visual reproduction was related more consistently and strongly to age than the two verbal subtests. The effects of selective attrition from the study were discussed.