Older women (mean = 69 yr) and younger women (mean = 23 yr) received 100 administrations of the digit symbol substitution subtest of the wechsler adult intelligence scale, modified so that the digitsymbol keys differed on each of the five training days. The older women improved substantially both within and between training sessions, but equally large gains were made by the younger women so that age differences were not reduced by training. As a consequence of practice, scores of the older women on the standard wais subtest increased from the 25th to the 90th percentile relative to norms for young adults, and these scores were maintained during a follow-up test 10 days later. Although practice effects transferred to digit-symbol keys different from the training keys, improvements did not generalize to WAIS digit span performance or to a matching task containing some of the elements of the training task.