The effectiveness of various sized shunts placed between the ascending and the descending aorta to prevent paraplegia in dogs with the thoracic aorta cross-clamped for 1 hour was tested. Three tapered shunts sizes were used with tip dimensions of 3.8, 5.2, and 6.3 mm inner diameter, with cross-sectional areas of 11.34, 21.23, and 33.18 mm2, respectively, and with an equal midportion diameter of 10 mm (3/8 inch). These shunts carried 40%, 60%, and 72% respectively, of baseline descending aortic flow during the cross-clamping period. Flow distribution was measured with radioactive microspheres in the spinal cord (gray and white matter) and kidneys. All dogs without shunts (Group I) developed paraplegia, severe proximal circulatory embarrassment, and severe ischemia of the spinal cord (mainly gray matter) that was followed by marked hyperemia persisting up to 24 hours following the experiment. Mortality was 33%. Only animals treated with large shunts (Groups III and IV) avoided paraplegia and postischemic injury. An effective shunt was characterized as carrying 60% or more of baseline descending aortic flow, having a cross-sectional area at its tip equal to or larger than 29% of the descending aorta, and equaling at least 54% of its diameter. Porportionately, the size of the tridodecylmethylamonium-heparin shunts being used in human beings (even the largest 9 mm inner diameter) is significantly inadequate to maintain distal flows and pressures for the prevention of spinal cord injury. Four clinical options are discussed.