We analyzed the cases of seventy-three patients who were managed over a twenty-year period for rheumatoid involvement of the cervical spine and were followed for a minimum of two years, with an average follow-up of seven years. A neurological deficit did not develop in thirty-one patients (Ranawat et al. Class I) and paralysis developed in the remaining forty-two patients: Class II in eleven and Class III in thirty-one. Of the forty-two patients in whom paralysis developed, thirty-five had operative stabilization. Seven patients were managed with a soft cervical collar because they refused or were medically unable to have the operation; all of the had an increase in the severity of the paralysis. The posterior atlanto-odontoid interval and the diameter of the subaxial sagittal canal measured on the cervical radiographs demonstrated statistically significant correlations with the presence and severity of paralysis. All of the patients who had a Class-III neurological deficit had a posterior atlanto-odontoid interval or diameter of the subaxial canal that was less than fourteen millimeters. In contrast, the anterior atlanto-odontoid interval, which has traditionally been reported, did not correlate with paralysis. The prognosis for neurological recovery following the operation was not affected by the duration of the paralysis but was influenced by the severity of the paralysis at the time of the operation. The most important predictor of the potential for neurological recovery after the operation was the preoperative posterior atlanto-odontoid interval. In patients who had paralysis due to atlanto-axial subluxation, no recovery occurred if the posterior atlanto-odontoid interval was less than ten millimeters, whereas recovery of at least one neurological class always occurred when the posterior atlanto-odontoid interval was at least ten millimeters. If basilar invagination was superimposed, clinically important neurological recovery occurred only when the posterior atlanto-odontoid interval was at least thirteen millimeters. All patients who had paralysis and a posterior atlanto-odontoid interval or diameter of the subaxial canal of fourteen millimeters had complete motor recovery after the operation. In this series, although only patients who had a neurological deficit were operated on, we observed the range of the posterior atlanto-odontoid interval that was associated with poor or no recovery after the operation, and we identified the safe range on the basis of the patients in whom paralysis did not develop.(ABSTRACT TRUNCATED AT 400 WORDS)