The hypothesis that collateral sprouting might be correlated with recovery of function was tested by a combination of behavioral and anatomical studies. Partial hemisections, sparing the dorsal columns, were made between T l r and L1 of cat spinal cord. Initially, all reflex and locomotor hindlimb activity was depressed ipsilaterally. During two postoperative weeks behavioral and electromyographic responses increased in response to intrinsic reflex elicitation but not to descending or crossed reflex elicitation. Marked improvement in use of the limb occurred as intrinsic reflex activity increased and became, in some cases, hyperactive. The status then remained stable for ten months. Since intrinsic reflexes are mediated by ipsilateral dorsal roots, the intraspinal distribution of L5 or L,; dorsal roots was determined by degeneration methods ten months after hemisection or by quantitative radioautography 20 days after hemisection. By both methods, the dorsal root distribution was found to be asymmetrical in the spinal gray. It was greater on the experimental side and the distribution was altered, having a greater distribution medially in Rexed's lamina VI and laterally in lamina VII. These anatomical changes are considered as signs of collateral sprouting from dorsal roots in response to degeneration of descending tracts on the same side. Dorsal root distribution to dorsal horn (laminae I-IV), motor nuclei, and nucleus gracilis is symmetrical. Regions of increased and of expanded dorsal root input can be correlated with electrophysiologically determined location of interneurons in the path of cutaneous reflexes and of stretch reflex facilitation. Behavioral and anatomical changes were selective; intrinsic reflexes recovered or became hyperactive and ipsilateral dorsal roots showed evidence of collateral sprouting during the recovery period, although a causal relationship between the two cannot be proven, by the present experiments.