Objective. To systematically review the measurement properties (i.e. internal consistency, reproducibility, validity, responsiveness and interpretability) of all performance-based methods which have been used to measure the physical function of patients with osteoarthritis of the hip or knee. Methods. A systematic search was conducted in Medline, CIN AHL, PsychINFO and Embase. Standardized criteria were applied to assess the quality of the clinimetric studies and the measurement properties. Results. Twenty-six performance-based methods were included: 13 walking tests, two stair-climb tests, one chair test and ten multi-item tests. Three out of seven multi-activity tests were tested for internal consistency and two were rated positively. Fourteen tests were tested for reliability and five were rated positively. The absolute measurement error (agreement) was assessed for 10 tests. Only one test received a positive rating. Fourteen tests were tested for construct validity. Only two tests received positive ratings. Responsiveness was assessed for 12 tests, but none of them received a positive rating. A lot of indeterminate ratings were given, mostly for small studies or non-optimal analyses. Conclusion. Many more well-designed studies are needed to assess the measurement properties of performance-based methods. More importantly, however, before one can make a justified choice of a particular performance-based method, consensus is needed on what activities should be included in a performance-based test for patients with hip or knee osteoarthritis and which aspects of function should be measured.