Amongst a series of 332 patients undergoing a prospective evaluation o
f the influence of magnetic resonance imaging (MRI) on diagnosis and o
utcome, arthroscopic correlation became available in 92 knees (91 pati
ents). The MRI report, the arthroscopic findings and the patient's cas
e notes were reviewed by a consensus panel. With respect to the menisc
i and cruciate ligaments, the formal radiological report was at varian
ce with the arthroscopic findings in 22/349 sites where there had been
no previous surgery (10 medial meniscus, six lateral meniscus and six
anterior cruciate ligament). 12 of these 22 errors were considered, o
n review, to be genuine MRI errors. However, considerable controversy
remains as to whether the other 10 ''errors'' were true or false. Ther
e are several factors which contribute to such controversy. These incl
ude technically difficult arthroscopies, delays between MRI and arthro
scopy, and ambiguities in the wording of both the referral letter and
the radiological report. These factors should be considered when evalu
ating the diagnostic performance of both MRI and arthroscopy.