In a prospective randomized study, we evaluated the risk of lumbar posterolateral spinal fusion at an unintended level due to bone graft migration. 130 patients underwent fusion supplemented by pedicle screw fixation (Cotrell-Dubousset, 64 patients) or uninstrumented fusion (66 patients). This was assessed by two independent observers on antero-posterior, and lateral radiographs taken 1 year after surgery. All patients had ben operated on at the preoperatively planned levels. Both observers agreed that fusion had taken place at an unintended level in 19 cases (14%). We found a tendency towards a higher risk of this "complication" when using supplementary pedicle screw fixation. The functional outcome, assessed by the Dallas Pain Questionnaire and the Low Back Pain Rating scale, was similar in patients having fusion at an unintended level and in patients fused only at the intended levels. There was no difference between the two groups concerning reoperation rates, postoperative smoking or social status. We conclude that unintended fusion occurs and tends to be commoner with the use of pedicle screw instrumentation. However, this complication seems not to affect the functional outcome if fusion has taken place at the intended level.