Background: This study examines the association of both pain severity and within-person pain variability with physical activity (PA) in older adults with osteoarthritis (OA). Methods: Data from the European Project on OSteoArthritis were used. At baseline, clinical classification criteria of the American College of Rheumatology were used to diagnose OA in older adults (65-85 years). At baseline and 12-18 months follow-up, frequency and duration of participation in the activities walking, cycling, gardening, light and heavy household tasks, and sports activities were assessed with the Longitudinal Aging Study Amsterdam Physical Activity Questionnaire. Physical activity was calculated in kcal/day, based on frequency, duration, body weight and the metabolic equivalent of each activity performed. At baseline and 12-18 months follow-up, pain severity was assessed using the pain subscales of the Western Ontario and McMaster Universities OA Index and the Australian/Canadian Hand OA Index. Within-person pain variability was assessed using two-week pain calendars that were completed at baseline, 6 months follow-up and 12-18 months follow-up. Results: Of all 669 participants, 70.0% were women. Sex-stratified multiple linear regression analyses showed that greater pain severity at baseline was cross-sectionally associated with less PA in women (Ratio = 0.95, 95% CI = 0.90-0.99), but not in men (Ratio = 0.99, 95% CI = 0.85-1.15). The longitudinal analyses showed a statistically significant inverse association between pain severity at baseline and PA at follow-up in women (Ratio = 0.94, 95% CI = 0.89-0.99), but not in men (Ratio = 1.00, 95% CI = 0.87-1.11). Greater pain variability over 12-18 months was associated with more PA at follow-up in men (Ratio = 1.18, 95% CI = 1.01-1.38), but not in women (Ratio = 0.94, 95% CI = 0.86-1.03). Conclusions: Greater pain severity and less pain variability are associated with less PA in older adults with OA. These associations are different for men and women. The observed sex differences in the various associations should be studied in more detail and need replication in future research.