Group I afferents in nerves innervating the lateral gastrocnemius-soleus (LG-Sol), plantaris (Pl), and vastus lateralis/intermedius (VL/VI) muscles were stimulated during walking in decerebrate cats. The stimulus trains were triggered at a fixed delay following the onset of bursts in the medial gastrocnemius muscle. Stimulation of all three nerves with long stimulus trains (> 600 ms) prolonged the extensor bursts and delayed the onset of flexor burst activity. LG-Sol nerve stimulation had the strongest effect; often delaying the onset of flex- or burst activity until the stimulus train was ended. By contrast, flexor bursts were usually initiated before the end of the stimulus train to the Pl and VL/VI nerves. The minimum stimulus strength required to increase the cycle period was between 1.3 x threshold and 1.6 x threshold for all three nerves. Simultaneous stimulation of the Pl and VL/VI nerves produced a larger effect on the cycle period than stimulation of either nerve alone. The spatial summation of inputs from knee and ankle muscles suggests that the excitatory action of the group I afferents during the stance phase is distributed to all leg extensor muscles. Stimulation of the group I afferents in extensor nerves generally produced an increase in the amplitude of the heteronymous extensor EMG towards the end of the stance phase. This increase in amplitude occurred even though there were only weak monosynaptic connections between the stimulated afferents and the motoneurones that innervated these heteronymous muscles. This suggests that the excitation was produced via oligosynaptic projections onto the extensor motoneuronal pool. Stimulation with 300 ms trains during the early part of flexion resulted in abrupt termination of the swing phase and reinitiation of the stance phase of the step cycle. The swing phase resumed coincidently with the stimulus offset. Usually, stimulation of two extensor nerves at group I strengths was required to elicit this effect. We were unable to establish the relative contributions of input from the group Ia and group Ib afferents to prolonging the stance phase. However, we consider it likely that group Ib afferents contribute significantly, since their activation has been shown to prolong extensor burst activity in reduced spinal preparations. Thus, our results add support to the hypothesis that unloading of the hindlimb during late stance is a necessary condition for the initiation of the swing phase in walking animals.