Phase-dependent reflex modulation was studied by recording the electromyographic (EMG) responses in ankle flexors (Tibialis Anterior, TA) and extensors (Gastrocnemius Medialis, GM and Soleus, SOL) to a 20 ms train of electrical pulses, applied to the tibial or sural nerve at the ankle, in human volunteers walking on a treadmill at 4 km/h. For low intensity stimuli (i.e. 1.6 times perception threshold), given during the swing phase, the most common response was a suppression of the TA activity with a latency of 67 to 118 ms. With high intensity of stimulation (i.e. 2.8 x T), a facilitatory response appeared in TA with a latency of 74 ms. This latter response was largest during the middle of the swing phase, when it was correlated with exaggerated ankle dorsiflexion. The TA reflex amplitude was not a simple function of the level of spontaneous ongoing activity. During stance, TA responses were small or absent and accompanied by a suppression of the GM activity with a latency ranging from 62 to 101 ms. A few subjects showed an early facilitatory, instead of a suppressive, GM response (88 to 136 ms latency). They showed a phase-dependent reflex reversal from a dominant TA response during swing to a facilitatory GM response with an equivalent latency during stance. The GM facilitation occurred exclusively during the early stance phase and habituated more than the TA responses. It is concluded that phase-dependent gating of reflexes occurs in ankle muscles of man, but only when vigorous extensor reflexes are present. More commonly, a phase-dependent modulation is seen, both of facilitatory and suppressive responses.