Data on articular cartilage thickness in the living are important for the design for computer models, aimed at preoperatively assessing the effect of surgical procedures on joint contact and load transmission, and for the calculation of cartilage material properties from its deformational behavior as determined during arthroscopy. A non-invasive method for measuring cartilage thickness in living subjects is, however, not available. A technique based on magnetic resonance imaging has therefore been tested for assessing articular cartilage thickness throughout joint surfaces. The accuracy is determined by comparing cartilage thickness maps obtained from three patellar specimens with a fat-suppressed three-dimensional gradient-echo sequence (resolution 2 x 0.31 x 0.31 mm) to those obtained with CT arthrography, A-mode ultrasound and anatomical sections. The distribution patterns are quantitatively compared using image analysis. The highest agreement was obtained for MRI and the sections (50% identical pixels), but all techniques yielded very similar results. On average, MR slightly underestimated the cartilage thickness compared with CT and the sections, and overestimated it compared with ultrasound. No evidence of differences in the degree of similarity could be detected in areas of thin and thick cartilage. We conclude that, if the resolution and accuracy of the method presented are considered acceptable, MRI is available for repeatable determination of topographical maps of articular cartilage thickness in living subjects. These data can be used to determine joint contact and stress in computer models and to calculate cartilage material properties in vivo.