This prospective, randomised study compares the efficacy of surgery, physiotherapy and cervical collar with respect to pain, motor weakness and sensory loss in 81 patients with long-lasting cervical radiculopathy corresponding to a nerve root that was significantly compressed by spondylotic encroachment, with or without an additional bulging disk, as verified by MRI or CT-myelography. Pain intensity was registered on a visual analogue scale (VAS), muscle strength was measured by a hand-held dynamometer, Vigorometer and pinchometer. Sensory loss and paraesthesia were recorded. The measurements were performed before treatment (control 1), 4 months after the start of treatment (control 2) and after a further 12 months (control 3). A healthy control group was used for comparison and to test the reliability of the muscle-strength measurements. The study found that before start of treatment the groups were uniform with respect to pain, motor weakness and sensory loss. At control 2 the surgery group reported less pain, less sensory loss and had better muscle strength, measured as the ratio of the affected side to the non-affected side, compared to the two conservative treatment groups. After a further year (control 3), there were no differences in pain intensity, sensory loss or paraesthesia between the groups. An improvement in muscle strengths, measured as the ratio of the affected to the non-affected side, was seen in the surgery group compared to the physiotherapy group in wrist extension, elbow extension, shoulder abduction and internal rotation, but there were no differences in the ratios between the collar group and the other treatment groups. With respect to absolute muscle strength of the affected sides, there were no differences at control 1. At control 2, the surgery group performed somewhat better than the two other groups but at control 3 there were no differences between the groups. We conclude that pain intensity, muscle weakness and sensory loss can be expected to improve within a few months after surgery, while slow improvement with conservative treatments and recurrent symptoms in the surgery group make the 1-year results about equal.