Cardiorespiratory responses to four patterns of arm-crank training (thrice weekly sessions at 50 or 70% of peak oxygen intake, 20 or 40 min per session) were examined over 8, 16, and 24 weeks in 24 initially inactive subjects with paraplegia. Training was associated with a significant increment of the peak oxygen intake during arm-crank tests except in control subjects and those combining a low intensity (50% of peak) with short-duration training (20-min sessions). There were associated increases in cardiac stroke volume, as assessed by a carbon dioxide rebreathing technique during submaximal exercise. It is suggested that the performance of inactive wheelchair users is limited by a pooling of blood in paralysed regions, with a reduction of cardiac preloading; nevertheless, substantial gains of performance are possible through prolonged arm-crank exercise of moderate intensity.