Reliable animal models of spinal cord injury are essential for studying pathological mechanisms and for laboratory testing of experimental treatments. The normal unpredictability of neurological outcome following experimental injury results partly from variations in the mechanics of both apparatus and tissue. Weight drop contusion models have been used extensively, and often effectively within a given study, but direct comparison between studies is usually made impossible by differences in the experimental parameters. The most important differences include the weight-height combination, the mass of the interface between weight and cord, and the support given to the cord from below. There are also important dimensional and physiological variables intrinsic to the biological material, which are usually ignored. A morphometric study of contusion injuries of the cat thoracic cord indicates that the major determinant of axon disruption is the extrusion of tissue from the impact site, due to viscoelastic distortion of the parenchyma within the meningeal tube. Direct compression and shear do not appear to play an important role in this kind of injury, where a brief compression of the cord occurs at an initial velocity of about 1.5 m/sec. The pathology produced by slower compression rates may vary, but the pattern of central necrosis, expected to be produced by extrusion, is common to most types of experimental lesion and to a large proportion of human injuries.