Objective: Previous studies of handedness and cognitive function rely on self-classification and yield inconsistent results. This study examines the associations of self-reported versus grip-strength-based handedness with cognitive function in healthy older men and women.Design: Cross-sectional study.Setting: 1988–91 follow-up clinic visit and 1991 mailed survey.Participants: 684 men and 985 women aged 55–95 who were community dwelling.Measurements: Cognitive function was assessed with 12 tests and grip strength was measured by hand-held dynamometer. Self-reported handedness was obtained with a mailed survey.Results: By self-report, 92.1% of men and women were right-handed; 2.0% were left handed. By grip strength, in men, 64.3% were right-handed, 22.5% left-handed, and 13.2% ambidextrous. In women, 61.3% were right-handed, 17.3% left-handed, and 21.4% ambidextrous. No cognitive function differences were found by self-reported handedness in either sex (p's>0.10). However, based on grip strength, left-handed women scored poorer than right-handed or ambidextrous women in immediate and delayed memory, attention, and verbal fluency (p's<0.05). Using categorical definitions, left-handed or ambidextrous individuals based on grip strength were more likely to show poor cognitive function on 4 of 5 tests.Conclusion: Grip strength is a useful alternative to self-reports for classifying handedness. Left-handedness by grip-strength, may be related to poorer cognitive function; this association may vary by gender.