OBJECTIVES: To analyze the effects of motor learning on knee extension-flexion isokinetic performance in knee osteoarthritis patients. METHODS: One hundred and thirty-six middle-aged and older sedentary individuals (111 women, 64.3±9.9 years) with knee osteoarthritis (130 patients with bilateral) and who had never performed isokinetic testing underwent two bilateral knee extension-flexion (concentric-concentric) isokinetic evaluations (5 repetitions) at 60°/sec. The tests were first performed on the dominant leg with 2 min of recovery between test, and following a standardized warm-up that included 3 submaximal isokinetic repetitions. The same procedure was repeated on the non-dominant leg. The peak torque, peak torque adjusted for the body weight, total work, coefficient of variation and agonist/antagonist ratio were compared between tests. RESULTS: Patients showed significant improvements in test 2 compared to test 1, including higher levels of peak torque, peak torque adjusted for body weight and total work, as well as lower coefficients of variation. The agonist/antagonist relationship did not significantly change between tests. No significant differences were found between the right and left legs for all variables. CONCLUSION: The results suggest that performing two tests with a short recovery (2 min) between them could be used to reduce motor learning effects on clinical isokinetic testing of the knee joint in knee osteoarthritis patients.