Walking in spinal-cord-injured (SCI) subjects is usually achieved at a lower speed than in normal subjects. Study design/methods: Time and distance parameters, angular displacements of lower limbs and electromyographic (EMG) activity were measured for seven normal and seven SCI subjects at several walking speeds. Analyses of variance were used for comparing groups and speeds. Objectives: First, to measure the adaptability of SCI subjects' walking pattern to different speeds (0.1-1.0 m/s), and to compare it to that of normal subjects. Second, to characterize SCI subjects' walking pattern as compared to that of normal subjects at a matched treadmill speed (0.3 m/s). Setting: University-Based Human Gait Laboratory, Montreal, Canada. Results: SCI subjects' pattern adapted to a limited range of speeds. Longer cycle duration, flexed knee at foot contact, increased hip joint flexion at foot contact and during swing, and altered coordination of hip and knee joints were found for the SCI group. At all speeds, duration of muscle activity was longer in the SCI group and the increase in amplitude of soleus EMG activity at higher speeds was not specific to push-off. The importance of matching the walking speed of SCI and normal subjects in order to differentiate the features that are a consequence of SCI subjects' reduced walking speed rather than a direct consequence of the injury is demonstrated. Conclusions: All SCI subjects could adapt to a narrow range of speeds and only three could reach the maximal tested speed. This limited maximal speed seems to be a consequence of SCI subjects having reached their limit in increasing stride length and not being able to increase stride frequency further. This limitation in increasing stride frequency is likely because of the altered neural drive. Sponsorship: Neuroscience Network of the Canadian Centre of Excellence.