Motor evoked potential (MEP) amplitudes have the disadvantage of a high variability when repeatedly assessed. This affects the reliability of MEP amplitude measurements taken during the course of motor incomplete spinal cord injury (iSCI). The study investigated the reliability of anterior tibial (TA) MEP measures controlled for dorsal flexion torque and motor task. TA MEPs were recorded at 10, 20, 40 and 60% of maximal voluntary contraction (MVC) during a static and dynamic (isometric increase of dorsal flexion torque) motor task. To determine reliability, 20 healthy and five chronic iSCI subjects were tested twice (> or =7 days) by the same investigator. Intraclass correlation coefficients (ICCs) were calculated. MEP amplitudes and latencies were compared between 20 healthy and 29 iSCI subjects. The reliability of MEP amplitude was in general good (ICC > or = 0.52) and was highest during the static task at 40% MVC (ICC = 0.77). The increased facilitation by the dynamic motor task showed the best reliability at 20% MVC (ICC = 0.48). The reliability was good to excellent for MEP latency (0.46 < or = ICC < or = 0.81), MVC (ICC > or = 0.90) and for the TMS threshold required to evoke a MEP response (ICC > or = 0.77). The torque generated by the MEP response ()0.02 < or = ICC < or = 0.55) and the duration of the silent period (0.07 < or = ICC < or = 0.50) were not reliable. Both MEP amplitudes and latencies differed significantly between healthy and iSCI subjects. Controlling for torque generation and motor task establishes a reliability of TA MEP amplitudes that is sufficient for longitudinal assessments in motor incomplete SCI.