The authors tested the use of repeated cognitive measures to identify those older individuals who experience cognitive decline after administration of a centrally acting medication. Subjects were 30 medically stable, older volunteers, mean age 71.4. Administration was a double-blind, placebo-controlled, crossover "challenge" from baseline to placebo, 50 mg, or 75 mg of diphenhydramine. More patients exhibited significant cognitive decrements after 75 mg of diphenhydramine than after placebo on the Buschke Selective Reminding Test Total Recall and the Trails B test. Data were consistent with models based on prediction intervals. Quantitative algorithms using prediction equations can characterize the normal limits of within-subject variability and define excessive cognitive change in elderly subjects, showing promise for use in monitoring patients to identify those who experience mild cognitive toxicity from prescribed medications.