1. In human subjects, stretch applied to ankle dorsiflexors elicited three bursts of reflex activity in the tibialis anterior (TA) muscle (labelled M1, M2 and M3) at mean onset latencies of 44, 69 and 95 ms, respectively. The possibility that the later of these reflex bursts is mediated by a transcortical pathway was investigated. 2. The stretch evoked a cerebral potential recorded from the somatosensory cortex at a mean onset latency of 47 ms in nine subjects. In the same subjects a compound motor-evoked potential (MEP) in the TA muscle, evoked by magnetic stimulation of the motor cortex, had a mean onset latency of 32 ms. The M1 and the M2 reflexes thus had too short a latency to be caused by a transcortical pathway (minimum latency, 79 ms (47 + 32)), whereas the later part of the M2 and all of the M3 reflex had a sufficiently long latency. 3. When the transcranial magnetic stimulation was timed so that the MEP arrived in the TA muscle at the same time as the M1 or M2 reflexes, no extra increase in the potential was observed. However, when the MEP arrived at the same time as the M3 reflex a significant (P < 0.01) extra-facilitation was observed in all twelve subjects investigated. 4. Peaks evoked by transcranial magnetic stimulation in the post-stimulus time histogram of the discharge probability of single TA motor units (n = 28) were strongly facilitated when they occurred at the same time as the M3 response. This was not the case for the first peaks evoked by electrical transcranial stimulation in any of nine units investigated. 5. We suggest that these findings are explained by an increased cortical excitability following TA stretch and that this supports the hypothesis that the M3 response in the TA muscle is - at least partly - mediated by a transcortical reflex.