The time course of regeneration of supraspinal and descending brachial intraspinal axons was studied using HRP retrograde tracing and kinematic analysis. Five groups of salamanders (10 salamanders/group) received complete thoracic transection 1.0 cm rostral to the hind limbs abolishing swimming. Groups 1-4 recovered for 2, 4, 6, and 8 weeks, respectively, before being filmed to record the animal's ability to swim. After filming, a second transection was made 1.0 cm caudal to the first (at the level of the lumbar enlargement) and HRP was used to label descending axons which had grown past the first lesion. The fifth group was filmed every 2 weeks for 12 weeks before the second transection was made for HRP application. The films were used to perform frame by frame computer analysis of the amplitude and timing of cyclic lateral flexion waves which make up swimming behavior. The earliest return of coordinated swimming behavior was seen 4 weeks after transection (1 of 20 animals). At 6 weeks post-transection, 5 of 10 animals exhibited coordinated swimming. However, the behavior in these animals was subnormal. In the group surviving 8 weeks post-transection, 5 of 10 animals recovered coordinated swimming behavior. In the group that was filmed every 2 weeks, 5 of the 10 salamanders which did recover, exhibited coordinated swimming behavior by the eighth week post-transection. Kinematic analysis of salamanders that exhibited a return of coordinated swimming revealed quantitative differences compared to normal salamanders. While continuous head to tail undulatory waves were present, the propagation time and period were faster than those in normal salamanders. Retransection of the spinal cord abolished coordinated swimming. The numbers and distribution of HRP-labeled supraspinal neurons varied greatly among the animals that displayed recovery of locomotor abilities. In the salamanders examined 6 weeks post-transection the majority of labeled cells were found in medullary nuclei. In recovered salamanders examined 8 and 12 weeks post-transection, HRP-labeled neurons were found in the red nucleus, in the interstitial nucleus of the fasciculus longitudinalis medialis, and in the mesencephalic as well as the medullary reticular neurons. Recovery of coordinated swimming was only observed in salamanders in which descending supraspinal and intraspinal axons were present at the level of the lumbar enlargement (as demonstrated by HRP retrograde labeling). These results indicate that recovery of locomotion is dependent on the reestablishment of descending input and is not a result of changes in spinal reflexes or propagation of electrical activity through the body wall.