1. Short-latency responses of single motor units (SMUs) and surface electromyographic activity (EMG) to transcranial magnetic stimulation (TMS) were examined in five different hand and forearm muscles of human subjects. 2. The response probability, P (number of extra spikes in the response peak above background per stimulus), was, in general, higher at the lower voluntary discharge rate of the motor unit than at the higher rate. 3. Increasing the strength of TMS increased the response probability of a tonically firing motor unit and at the same time recruited new units which discharged phasically during the response peak. This demonstrates rate coding and recruitment of motor units by excitatory inputs resulting from TMS when the motoneurone pool is tonically facilitated by a constant voluntary drive. 4. Next, TMS was delivered without any voluntary facilitation of motoneurones. The order of recruitment for up to four different motor units discharged by TMS was compared to that observed with voluntary input. The threshold of recruitment for each of the two inputs was estimated from the surface EMG value at which the unit was recruited. For these motoneurone pools (eleven sets of observations), the order of recruitment was the same with TMS and voluntary inputs. 5. From these data it is concluded that, despite the complex and phasic nature of the descending corticospinal volleys generated by TMS, it produces orderly recruitment and rate coding of motoneurones similar to that found for voluntary activation.