In this investigation we have estimated the afferent contribution to the generation of activity in the knee and ankle extensor muscles during walking in decerebrate cats by loading and unloading extensor muscles, and by unilateral deafferentation of a hind leg. The total contribution of afferent feedback to extensor burst generation was estimated by allowing one hind leg to step into a hole in the treadmill belt on which the animal was walking. In the absence of ground support the level of activity in knee and ankle extensor muscles was reduced to approximately 70% of normal. Activity in the ankle extensors could be restored during the "foot-in-hole" trials by selectively resisting extension at the ankle. Thus feedback from proprioceptors in the ankle extensor muscles probably makes a large contribution to burst generation in these muscles during weight-bearing steps. Similarly, feedback from proprioceptors in knee extensor appears to contribute substantially to the activation of knee extensor muscles because unloading and loading these muscles, by lifting and dropping the hindquarters, strongly reduced and increased, respectively, the level of activity in the knee extensors. This conclusion was supported by the finding that partial deafferentation of one hind leg by transection of the L4-L6 dorsal roots reduced the level of activity in the knee extensors by approximately 50%, but did not noticeably influence the activity in ankle extensor muscles. However, extending the deafferentation to include the L7-S2 dorsal roots decreased the ankle extensor activity. We conclude that afferent feedback contributes to more than one-half of the input to knee and ankle extensor motoneurons during the stance phase of walking in decerebrate cats. The continuous contribution of afferent feedback to the generation of extensor activity could function to automatically adjust the intensity of activity to meet external demands.