In the present study we investigated the validity of magnetic resonance imaging (MRI) and arthroscopy in knees with acute, traumatic extension deficit (the "locked knee"), and evaluated whether arthroscopy of knees with no mechanical pathology could be avoided by MRI evaluation. The study consisted of 50 patients who had an acute, traumatic extension deficit of the knee. All patients were submitted to MRI prior to arthroscopy. Following MRI and surgery, standardized forms were filled out, attempting to objectify the findings. The orthopaedic surgeon was not aware of the MRI result prior to surgery. Evaluating MRI, all grade-3 meniscal lesions were considered able to cause a mechanical block as well as acute partial or total anterior cruciate ligament (ACL)-ruptures. ACL-ruptures with an old appearance were not considered able to cause locking. Assuming that arthroscopy was the gold standard, the following results were calculated for the overall appearance of a lesion able to cause locking: Positive predictive value = 0.85, negative predictive value = 0.77, sensitivity = 0.95, specificity = 0.53. Two knees were erroneously evaluated with no mechanical locking at MRI (one bucket-handle lesion and one pathological synovial plica). MRI of the knee with acute, traumatic extension deficit is in the acute or subacute phase a safe method to identify the patients that have a mechanical reason for locking and therefore can benefit from arthroscopic treatment.