An external device ("rottometer") specially designed to measure knee joint rotation was developed and evaluated with respect to its validity. Simultaneous measurements were made with the rottometer and Roentgen stereometric analysis (RSA) in five patients with implanted tantalum markers in the tibia and femur. Measurements of internal and external rotation were made at 90 degrees and 60 degrees of knee flexion using 3, 6 and 9 N m torques. The coefficients of determination (r2) between the results obtained with the rottometer and RSA were around 0.9 for the total rotation. The rottometer consistently overestimated the rotation by about 100% and this systematic error was most constant at 90 degrees flexion for the different torques. The magnitude of this error from soft tissue deformation as well as the rotatory movements in the hip, foot and ankle joints must be considered when using external devices to measure knee rotation in clinical studies. The most accurate registrations were found in 90 degrees flexion with 9 N m force (r2 = 0.94).