Quantitative assessment of cartilage volume and thickness in a formalin-alcohol fixed specimen of a human patella was conducted with magnetic resonance imaging (MRI), as it is still unclear whether the morphology of normal and damaged cartilage can be accurately demonstrated with this technique. MR imaging was carried out at 1.0 T (section thickness 2 mm, in-plane-resolution 0.39-0.58 mm) with the following pulse sequences: 1) T1-weighted spin-echo, 2) 3D-MPRAGE, 3) 3D-FISP, 4) 3D-MTC-FISP, 5) 3D-DESS, 6) 3D-FLASH. Following imaging, the patella was sectioned perpendicular to the articular surface at intervals of 2 mm with a diamond band-saw. The volume of its cartilage was determined from the anatomical sections and the MR images, using a Vidas IPS 10 image analysing system (Kontron). Measurements were carried out with and without the low-signal layer in the transitional zone between the articular cartilage and the subchondral bone. If the low-signal layer was included, the volume was overestimated with MRI by 16 to 19%. Without the low-signal layer the volumes were less than those determined from the anatomical sections: T1-SE-18.2%, MPRAGE -22.6%, FISP -17.1%, MTC-FISP -9.5%, DESS -9.3% and FLASH -6.1%. The coefficient of variation for a 6-fold determination of the volume amounted to between 6.2% (T1-SE) and 2.6% (FLASH). The FLASH sequence allowed the most valid and reproducible assessment of the cartilage morphology. The remaining difference from the real volume of the cartilage may be due to the fact that the calcified zone of the cartilage is not delineated by MRI.