We assessed the accuracy of clinical evaluation, arthrography, and arthroscopy in the diagnosis of meniscal lesions in fifty knees in which arthrotomy was performed for disabling symptoms after evaluation by these three methods. At surgery, forty-seven menisci were removed, of which forty-four were abnormal and three were normal. In three patients with normal menisci, loose bodies were found in two and the exploration was negative in one. In the forty-four knees with a meniscal lesion, a correct diagnosis was made clinically forty time, arthrographically thirty-nine times, and arthroscopically thirty-two times. Most errors occurred in the knees with posterior horn lesions of the medial meniscus. Clinical diagnosis was least accurate for lesions of the lateral meniscus (four missed) and arthroscopy was least accurate for lesions of the posterior horn of the medial meniscus (ten missed). Arthrography appeared to provide collateral evidence of lesions not seen directly. Based on this study it was concluded that even with negative findings by arthroscopy and arthrography it still may be necessary occasionally to remove a meniscus on the basis of the clinical evaluation.