A prospective, single-blind study of 57 consecutive knees with an initial clinical diagnosis of a torn meniscus was completed to test the hypothesis that a routine preoperative magnetic resonance imaging (MRI) scan could improve accuracy over clinical diagnosis alone. The overall accuracy for the clinical diagnosis of meniscal tear was 80.7% and the corresponding accuracy for MRI was 73.7%. Surgical pathology was found in all knees at arthroscopy. Relying blindly on MRI to determine surgical intervention would have resulted in inappropriate treatment in 35.1% of the knees. A surgeon can safely rely on the clinical diagnosis of meniscal tear alone and will find surgical pathology in almost all cases. Relying on MRI alone without using clinical judgment may lead to inappropriate treatment in a high percentage of cases. Clinical examination alone is at least as accurate as MRI. MRI did not prevent "unnecessary surgery" in any case. MRI should be reserved for confusing or special cases.