1. Previous studies indicate that after transection of the rostral spinal cord, larval lamprey begin to recover locomotor behavior 2 wk posttransection and recovery is complete at approximately 8 wk. To examine the mechanisms underlying behavioral recovery after spinal cord transection, in the present study the time course and extent of recovery of locomotor function was examined in in vitro brain/spinal cord preparations. With these preparations the contributions of functional regeneration of descending brain stem projections to recovery of spinal locomotor function can be examined in the absence of mechanosensory inputs and descending propriospinal relay systems. 2. In in vitro preparations from normal lamprey, stimulation in brain stem locomotor regions resulted in direct descending activation of locomotor networks in the rostral, middle, and caudal spinal cord. 3. At 4 wk posttransection, in vitro locomotor activity was usually confined to the rostral spinal cord a few millimeters below the transection site. At 8 wk posttransection locomotor activity was present in both the rostral and middle spinal cord, and spinal locomotor networks at these levels could be directly activated by restored descending projections from the brain stem. 4. At 16-32 wk posttransection locomotor activity similar to that seen in normal animals was present along the spinal cord. Additional manipulations suggest that at 32 wk posttransection descending axons from brain stem command/initiation neurons had grown for relatively long distances and could directly activate the locomotor networks in the caudal spinal cord. At each recovery time examined the ranges of locomotor parameters (cycle time, burst proportion, and intersegmental phase lag) overlapped with those in normal animals. 5. In vitro locomotor activity in spinal cord-transected animals could be recorded at progressively more caudal levels below the transection site during the course of recovery. However, locomotor activity in in vitro preparations occurred for shorter distances below the lesion than in whole animals at comparable recovery times. 6. Our recent double-labeling experiments suggest that behavioral recovery in spinal cord-transected lamprey is largely due to true regeneration of preexisting descending axons rather than development of new descending projections. Thus, these results in conjunction with our behavioral, in vitro, and anatomic data suggest that functional regeneration of descending axons from the brain, as well as other mechanisms such as descending propriospinal relay systems and mechanosensory inputs, account for the gradual restoration of locomotor function in spinal cord-transected lamprey.