We used the KT-1000 arthrometer to test the knees of 107 patients who had an acute tear of the anterior cruciate ligament, 153 patients who had a chronic tear, and 141 control subjects, for a total of 401 individuals. The three testing parameters were the extent of anterior translation at eighty-nine newtons of force and at maximum manual force, and the compliance index. The differences between the involved and the uninvolved knees were calculated. At eighty-nine newtons, all but one of the control subjects had anterior translation of ten millimeters or less, compared with 58 per cent of the patients who had a chronic tear. At maximum manual force, all but two of the control subjects had translation of ten millimeters or less, compared with 20 per cent of the patients who had an acute or a chronic tear. Analysis of variance showed that the clinical diagnosis correlated well with the results for all tests (p less than 0.001). However, when the uninjured knees of patients who had an acute or a chronic tear were compared with the knees of the control subjects, significant differences were noted (p less than 0.001 to 0.006). In the patients who had a chronic tear, there was no relationship between the time from injury to operation and the extent of anterior translation. The arthrometric test at maximum manual force was the strongest discriminant; it differentiated normal from abnormal knees (p less than 0.001) with high sensitivity (92 per cent), high specificity (95 per cent), and high positive predictive accuracy; the cut-off point was eleven millimeters or less.(ABSTRACT TRUNCATED AT 250 WORDS)