Arthroscopy of the temporomandibular joint (TMJ) is evaluated on an autopsy material. 2 different types of arthroscope--one with a rod-lens system and one with a so-called Selfoc system--were used and compared regarding diagnostic accuracy. On 54 cadavers, it was found that the upper joint compartment may be punctured with accuracy and without damage to vital tissues. Landmarks on the skin facilitating clinical arthroscopy were defined in relation to a guideline from tragus to lateral canthus. A further 49 TMJ specimens were investigated with arthroscopy. The findings are in good agreement with subsequent observations made during dissection. Regarding arthrotic changes, 100% diagnostic accuracy was achieved with both types of arthroscope. Regarding remodelling changes, the diagnostic accuracy was approximately 57%. No significant difference was noticed between the 2 arthroscopes. Photographic documentation with the rod-lens arthroscope was found to be of superior quality.