We evaluated motor unit (MU) fatigue in the first dorsal interosseous muscle (FDI) using the collision principle. Eight healthy men exerted 70% (short-duration fatigue task: SDK task) and 30% (long-duration fatigue task: LDF task) maximum voluntary contraction of isometric abductions in the left FDI until exhausted. Before and after voluntary contractions, the ulnar nerve was stimulated at the wrist and elbow with supramaximal intensity, and a pair of M-waves was obtained. Fatiguerelated changes were studied in mean power frequency (MPF), averaged rectified value (ARV) calculated from surface EMG, and motor nerve conduction velocity (MCV) and distribution of motor nerve conduction velocity (DMCV) calculated from M-waves. The MPF of voluntary EMG decreased, whereas ARV increased significantly during SDF and LDF tasks, indicating fatigue had developed in the FDI. Endurance was significantly shorter in the SDF task than in the LDF task (p<0.01), whereas differences between tasks were not seen in MPF and ARV changes. Tasks did not affect MCV, but lower components in DMCV increased for both tasks. Increased lower components were larger in the LDF task than in the SDF task. The shift in DMCV indicated that fatigued MUs stopped activity and enduring MUs, which had lower axon conduction velocity, were activated selectively. These results suggest that the collision principle is applicable in evaluating motor unit fatigability.