Between April and October 1981, nineteen children who were between five and twelve years old (median age, seven years), and who had severe paralytic scoliosis secondary to poliomyelitis, were treated by segmental spinal instrumentation without arthrodesis. Three patients could not be followed, and the cases of the remaining sixteen patients were reviewed in April 1984. In fifteen patients, the implant system had failed to control the deformity. In six patients, the rods had fractured at the apex of the original deformity; in five, longitudinal shift of the rods had allowed recurrence of the deformity; and in four, the short limb of one or both of the L-shaped rods had rotated out of the pelvis and perforated the skin, resulting in an infection that necessitated removal of the rods. At the time of the review, the deformity was essentially the same as before the original instrumentation in all but the sixteenth patient, in whom failure had not occurred. The instrumented vertebrae had grown measurably in some of the patients, but spontaneous fusion--particularly in the thoracic spine--was observed when the rods were removed. This finding indicates that, using presently available materials, segmental instrumentation of the spine without arthrodesis does not effectively control paralytic scoliosis secondary to poliomyelitis in a growing child.