BACKGROUND: Impaired joint position sense (JPS) has been shown in anterior cruciate ligament (ACL) deficient and osteoarthritic knees. The relation between JPS and function is uncertain. The aim of this study was to determine further if ACL deficient knees show abnormal JPS and the effect of exercise therapy on JPS, and also to assess the relation between JPS, functional stability, and strength. METHODS: Fifty patients (46 men and four women, mean age 26.3 years) with unilateral ACL deficient knees were assessed on admission and after rehabilitation (5 hours a day for four weeks). JPS was assessed by reproduction of passive positioning using a visual analogue incorporating a goniometer. Knee stability was analysed by self report questionnaire (score 0-280) and functional activity test (single leg hop and figure of eight run). Isokinetic dynamometry was performed to evaluate quadriceps and hamstring peak torque strength. Controls were either age and sex matched individuals or the contralateral knee. Statistical analysis was by Wilcoxon signed rank test and Spearman rank order correlation coefficient. RESULTS: JPS was impaired in ACL deficient knees. The mean (SD) errors in reproducing angles were 9.4 (3.1) degrees and 7.1 (2.3) degrees for the ACL deficient knee and control knee respectively (P < 0.0005). There was no improvement in JPS after rehabilitation (9.4 (3.1) degrees and 8.5 (3.2) degrees before and after rehabilitation respectively, P = 0.14). There was improvement as ascertained from the questionnaire (on admission 202 (32.1), after rehabilitation 243 (25.4), P < 0.0001) and functional activity testing (hop: on admission 148.7 (37.3) cm, after rehabilitation 169.8 (31.1) cm, P < 0.0005; figure of eight: on admission 48.4 (16.6) seconds, after rehabilitation 41.6 (3.4) seconds, P < 0.0001). Quadriceps strength improved (peak torque on admission 198.5 (58.9) Nm, after rehabilitation 210.5 (54.2) Nm, P < 0.05), but not hamstring strength (peak torque on admission 130.6 (28.1) Nm, after rehabilitation 135.5 (27.7) Nm, P = 0.24). JPS did not correlate with the functional activity tests (hop and figure of eight run), the responses to the questionnaire, or strength. There was no correlation between the responses to the questionnaire and functional activity tests or muscle strength. CONCLUSIONS: JPS was impaired in ACL deficient knees. Although knee stability improved with exercise therapy, there was no improvement in JPS. The role of JPS in the stability of ACL deficient knees remains unclear.