Mechanosensory neurons from a proprioceptor (the femoral chordotonal organ) signal the movements and positions of the femorotibial joint of a locust leg. Intracellular recordings from these neurons during walking show that their spikes are superimposed on a depolarizing synaptic input generated near their output terminals in the CNS. The depolarization consists of a rhythmic synaptic input at each step, and a sustained input that begins before walking commences. In different sensory neurons, which signal particular features of the movement, the rhythmic depolarization occurs at distinct times during either the swing or stance phases of the step cycle. The depolarizing input is timed to coincide with the greatest spike response of a sensory neuron. The input is associated with a conductance change, appears to reverse just above resting potential, and thus has similar properties to the presynaptic inhibition in these same neurons during imposed joint movements (Burrows and Laurent, 1993; Burrows and Matheson, 1994). Three sources could contribute to these inputs: (1) interactions between sensory neurons of the same receptor signaling the same movement, (2) signals from different receptors in the same leg and other legs, and (3) outputs of central neurons involved in generating walking. When the leg, whose movements the sensory neurons signal is removed, both the sustained and rhythmic synaptic inputs persist. Sensory neurons in isolated ganglia treated with pilocarpine are also depolarized in phase with a rhythmic pattern expressed in leg motor neurons, indicating that central neurons must contribute. The maintained synaptic input to the terminals means that the overall effectiveness of the sensory spikes in evoking EPSPs in postsynaptic neurons will be reduced during walking, and the rhythmic component means that the spikes from particular sensory neurons will be further reduced at particular phases of the step cycle that they signal best.