Percutaneous radiofrequency neurotomy has been used in the treatment of pain from the cervical zygapophysial joints, but the results have been modest and not compelling. Several factors might account for its apparent poor success rate, including inadequate patient selection, inaccurate surgical anatomy, and technical errors. In an effort to overcome these confounders, we used comparative local anesthetic blocks to preoperatively, definitively diagnose cervical zygapophysial joint pain and developed an amended operative technique based on formal anatomical studies. An audit was conducted of our experience with 19 patients to determine whether there was sufficient merit in the amended procedure to justify a randomized, double-blind, controlled trial. The duration of complete pain relief was the principal outcome measure. Side effects and complications were also monitored. Of the 10 patients who underwent third occipital neurotomy for the treatment of C2-C3 zygapophysial joint pain, only 4 obtained long-lasting relief. The other six patients reported an early return of their pain and constituted technical failures; the third occipital nerve was inadequately coagulated and recovered in the immediate postoperative period. Of the 10 patients who underwent lower cervical medial branch neurotomy, 7 obtained complete pain relief for clinically useful periods and were able to resume their activities of daily living and employment. After procedures at all levels, a brief period of postoperative pain was experienced by the patients and ataxia was a side effect of third occipital neurotomy. There were no cases of postoperative infection or anesthesia dolorosa. Given the high technical failure rate of third occipital neurotomy, we recommend that this procedure be abandoned until the technical problems can be overcome.(ABSTRACT TRUNCATED AT 250 WORDS)