1. Signal averaging was used in forty experiments on low-spinal cats to measure and characterize the oligosynaptic responses of seventy-six motoneurons supplying the medial gastrocnemius muscle to the single impulses of antidromically stimulated single motor axons supplying the same muscle. 2. In thirteen experiments on chloralose-urethane anaesthetized preparations, twelve (43%) of the tested twenty-eight motoneurones exhibited a single-axon recurrent inhibitory post-synaptic potential (recurrent i.p.s.p.), as compared to sixty-four (62%) of the 103 motoneurones tested in twenty-seven animals in the absence of anaesthetic after ischaemic decapitation. 3. Single-axon recurrent i.p.s.p.s most often consisted of a single, long-lasting hyperpolarization. Ten of the recurrent i.p.s.p.s contained a second late peak of hyperpolarization. In another eight of the i.p.s.p.s, a small late depolarization was evident. 4. The distinct profiles of the recurrent i.p.s.p.s were readily distinguished from the relatively flat profiles with low noise levels in the averages of the fifty-five 'no-response' cells. The transmembrane and post-synaptic nature of the i.p.s.p.s was confirmed by extracellular control recordings taken immediately outside seven of the cells with positive responses. In addition, ten cells with positive responses were subjected to current passage during the averaging procedure. In all cases, depolarization increased and hyperpolarization reduced the amplitude of their single-axon recurrent i.p.s.p.s. 5. The mean amplitude of the responses was 12.0 microV in chloralose-urethane preparations as compared to a peak-to-peak noise level less than 6.0 microV in the no-response averages. Corresponding values in ischaemic-decapitate preparations were 46.2 microV and less than 7.5 microV, respectively. 6. Latency, rise-time and half-width (i.e. duration at half-amplitude) values of the i.p.s.p.s were similar for chloralose-urethane and ischaemic-decapitate preparations. The average values in both preparations were 2.5, 5.6 and 19.3 ms, respectively. The latency values indicated both disynaptic and, perhaps, longer components in the recurrent i.p.s.p.s. The rise-time and half-width values were relatively similar to those reported or measured from published records for analogous composite recurrent i.p.s.p.s (i.e. responses to antidromic stimulation of the whole muscle nerve rather than single motor axons). A weak, but significant, correlation between rise-time and half-width was observed for the sixty-six single-axon recurrent i.p.s.p.s with a single negative-going peak.(ABSTRACT TRUNCATED AT 400 WORDS)