Thirty-nine patients with split cord malformations (SCM) were studied in detail with respect to their clinical, radiographic, and surgical findings as well as their outcome data. Eight patients were adults and 31 patients were children. According to the classification endorsed by Part I of the SCM study, 19 patients had Type I SCM (6 adults and 13 children), 18 patients had Type II SCM (2 adults and 16 children), and 2 patients had composite SCM with both lesion types situated in tandem. Six SCMs were cervical, 2 were thoracic, and 31 were in the lumbar region. All 8 adults had pain and progressive sensorimotor deficits at diagnosis. Only 16 of the 31 children had symptoms, and among these, 14 had progressive sensorimotor deficits, but only 6 had pain. The difference in the clinical picture between adults and children is similar to that described in the tethered cord syndrome, except for left-right functional discrepancy, which was prominent in 8 children with SCM but rarely seen in tethered cord syndrome due to other causes. Cutaneous manifestations of either occult or open dysraphic states were present in all but 3 patients; hypertrichosis was by far the best predictor of an underlying SCM, being found in 56% in the series. Neurological deterioration in SCM was independent of the lesion type: the Type I:Type II ratio for symptomatic progression was 13:11. It was also independent of the location of the lesion: 67% of patients with cervical SCMs had symptomatic progression versus 64% of patients with thoracolumbar lesions. High-resolution, thin cut, axial computed tomographic myelography using bone algorithms was more sensitive than magnetic resonance imaging in defining the anatomical details of the SCM. Radiographic classifications of the SCM, using the nature of the median septum and the number of dural tubes as criteria, was always possible without ambiguity. However, whereas every Type I bone septum was identified preoperatively, only 5 Type II fibrous septa were revealed by preoperative imaging, even though a fibrous septum and/or other fibroneurovascular bands were found tethering the hemicords in every Type II case at surgery. Complete imaging studies also showed that all lumbar SCMs had low-lying coni and at least one additional tethering lesion besides the split cords, whereas only 1 of 7 cervical and high thoracic SCMs had a low conus and a second tethering lesion. The surgical goal for SCM was release of the tethered hemicords by eliminating the bone spurs, dural sleeves, fibrous septa, or any fibroneurovascular bands (myelomeningoceles manqué) that might be transfixing the split cord. Type I cases were technically more difficult and had a slightly higher surgical morbidity than Type II cases, especially if an oblique bone septum had asymmetrically divided the cord into one larger hemicord and one smaller, hence, very delicate, hemicord.(ABSTRACT TRUNCATED AT 400 WORDS)