Functional magnetic resonance imaging (fMRI) has emerged during the last decade as the main non-invasive technique for the investigation of human brain function. More recently, fMRI was also proposed for functional studies of the human spinal cord, but with controversial results. In fact, the functional contrast is not well-characterized, and even its origin has been challenged. In the present work, we characterized the temporal features of the functional signal evoked in the human spinal cord by a motor task, studied with an approach based on time-locked averaging of functional time series of different durations. Based on the results here reported, we defined an impulse-response function (irf) able to explain the functional response for motor tasks in the interval of 15-42 s of duration, thus suggesting the linearity of the phenomenon in this interval. Conversely, with stimulation durations ranging between 3 and 9 s, the functional signal was not detectable, and was under the level predicted by a linear behavior, suggesting deviation from linearity during short stimulations. The impulse-response function appeared slower than in the brain, peaking at about 9 s after its beginning. The observed contrast was generally larger than in the brain, on the order of about 5.4% of baseline signal at 1.5 T. The findings further suggested that the physiological origin of T(2) weighted functional imaging is similar in the spinal cord and in the brain.