During a sustained maximal effort a progressive decline in the ability to drive motoneurones (MNs) develops. We used the recently developed triple stimulation technique (TST) to study corticospinal conduction after fatiguing exercise in healthy subjects. This method employs a collision technique to estimate the proportion of motor units activated by a transcranial magnetic stimulus. Following a sustained contraction of the abductor digiti minimi muscle at 50 % maximal force maintained to exhaustion there was an immediate reduction of the TST response from > 95 % to about 60 %. This effect recovered to control levels within 1 min and implies that a decreased number of spinal MNs were excited. Additional TST experiments after maximal and submaximal efforts showed that the decrease in size of the TST response was related to duration and strength of exercise. Motor evoked potentials (MEPs) after conventional transcranial magnetic stimulation (TMS) and responses to peripheral nerve stimulation were recorded following the same fatigue protocol. The size of both the MEPs and the peripheral responses increased after the contraction and were in direct contrast to the decrease in size of the TST response. This points to increased probability of repetitive spinal MN activation during fatigue even if some MNs in the pool failed to discharge. Silent period duration following cortical stimulation lengthened by an average of 55 ms after the contraction and recovered within a time course similar to that of the TST response depression. Overall, the results suggest that the outflow from the motor cortex could become insufficient to drive all spinal MNs to discharge when the muscle is fatigued and that complex interactions between failure of activation and compensatory mechanisms to maintain motor unit activation occur during sustained voluntary activity. When inability to maintain force occurs during submaximal effort, failure of activation of motor units is predominant.