To illuminate the action of non-invasive stimuli on the human cerebral cortex, responses of corticospinal axons and of plantar alpha-motor neurons following transcranial magnetic (TMS) and electrical stimulation (TES) were recorded in the lumbosacral cord in the anaesthetized macaque monkey. A round coil was used for TMS, and the anode was located at the vertex for TES. The responses of 175 identified corticospinal axons (conduction velocities of 24-95 m/s) were recorded from the lateral corticospinal tract at the T12-L3 spinal level. A single magnetic or electrical stimulus could evoke an early spike corresponding to the direct (D) wave in surface recorded volleys and was termed a D response. In the same axon, up to four further spikes, termed indirect (I) responses, could also be evoked. At a given intensity of stimulation, D responses had clear thresholds and fixed latencies, whereas I responses were labile in both respects. For TMS and TES, the thresholds of both D and I responses were inversely correlated with axonal conduction velocity. For TMS, fast conducting axons (> 75 m/s) had lower thresholds for D responses, while more slowly conducting axons (< 55 m/s) had lower thresholds for I responses. Very few of the axons with a conduction velocity of < 40 m/s (three out of 23) gave a D response to TMS. For TES, the majority of axons had lower thresholds for D responses or a similar threshold for both D and I responses. At threshold, the latencies of D responses evoked by TMS and TES were consistent with activation within the cortex, while TES also excited some corticospinal axons deep to the cortex. At 2.5 times threshold for the D response, TMS still excited axons mostly within the cortex, but with TES the site of activation shifted by as much as 65 mm below the cortex (mode 20 mm). Intracellular responses were recorded in 23 plantar alpha motor neurons supplying intrinsic muscles of the foot. All showed monosynaptic excitatory post-synaptic potentials (EPSPs) to both TMS and TES with no significant differences in the rise times of the evoked EPSPs. At threshold for a surface corticospinal volley, the average EPSP to TES began 0.5 ms earlier than that to TMS, and 1.0 ms earlier at 2.5 times this threshold. The different sites of activation of corticospinal neurons by TMS and TES, as well as the different distribution of D and I responses that they evoke, may both contribute to the differences in the onset latencies of the EMG responses evoked by these methods in human subjects.