One hundred and twenty-seven patients who had a rupture of the anterior cruciate ligament agreed to participate in a prospective, randomized study of three arthroscopically assisted reconstruction techniques. One hundred and twenty-five patients (125 reconstructions) were evaluated after a mean duration of follow-up of forty-two months (range, two to five years). Group I included forty patients who had a two-incision reconstruction with use of an autogenous semitendinosus-gracilis graft, group II consisted of forty patients who had a two-incision reconstruction with use of an autogenous patellar-ligament graft, and group III included forty-five patients who had a single-incision reconstruction (endoscopic technique) with use of an autogenous patellar-ligament graft. The male-female ratio, age range, level of athletic activity, interval between the injury and the reconstruction, previous operative procedures, and associated injuries were similar in all three groups. The same postoperative rehabilitation protocol was followed for all patients. Testing with a KT-2000 arthrometer at maximum manual force was done at the follow-up evaluation; the difference in laxity between the involved knee and the contralateral knee was three millimeters or less in thirty-three patients (83 per cent) in group I, thirty-seven patients (93 per cent) in group II, and thirty-nine patients (87 per cent) in group III. A difference of two millimeters or less was found in thirty patients (75 per cent) in group I, thirty-one patients (78 per cent) in group II, and thirty-five patients (78 per cent) in group III. Thirty-five patients (88 per cent) in group I, thirty-eight patients (95 per cent) in group II, and forty patients (89 per cent) in group III returned to at least the same level of athletic activity. Four grafts (two in group I and two in group II) failed as a result of trauma. There was one additional failure in groups I and III, as evidenced by a difference of nine and seven millimeters, respectively, on instrumented testing of laxity. The significant findings were that no knee was rated D according to the system of the International Knee Documentation Committee (p < 0.002, 94 per cent confidence level) and that fewer additional operative procedures were done on patients in group III (p < 0.08). Also, it was found that the patients in group II returned to a greater level of athletic activity (p < 0.02) and that a higher percentage of the patients in this group had a difference of three millimeters or less on testing with the KT-2000 arthrometer than in the other two groups (p < 0.08). However, with the numbers available, there were no significant differences in the over-all outcome among the three groups (p > 0.1). Importantly, the rate of failure was not greater and the outcomes were not less satisfactory for the late reconstructions than they were for the acute reconstructions (those performed less than three weeks after the injury), including those done with an autogenous semitendinosus-gracilis graft in a chronically unstable knee.