Previously reported experimental models for spinal cord contusion injury do not allow the independent control of compression and contact velocity required for interpretation of experimental data relating kinematics of vertebral injury to spinal cord injury. Therefore, controlled dynamic compression of the spinal cord was used to study compression and contact velocity as independent variables. Cord conduction was assessed using the latency of somatosensory evoked potentials in response to hindlimb stimulation. The latency increase at 4 hours after contusion differed significantly between control and 50% compression results, and between 25% and 50% compression results. A small nonsignificant increase in latency was observed with increase in contact velocity. The extent of hemorrhagic necrosis correlated with contact velocity rather than with the amount of compression. This study demonstrates, for the first time, a dissociation between hemorrhagic necrosis and loss of neuronal conduction in the subacute phase. Although long-term effects of hemorrhagic necrosis on cord structure and conduction remain to be evaluated, the data suggest that delayed loss of neuronal conduction seen clinically may result from moderate levels of cord compression at high contact velocity. Such an injury is not reproducible by weight-drop techniques for cord injury.