A model of spinal cord trauma in guinea pigs is described, based on the concept of compression to a set thickness, as an alternative to compression or contusion with a set force or displacement. The model is technically simple and reliable and circumvents some of the biomechanical problems of contusion techniques. It was designed initially to produce moderate injuries, allowing significant recovery of function. A pair of forceps was modified to form an instrument to compress the spinal cord laterally, over a 5-mm length, to a thickness of 1.2 mm. Such compression injuries of the lower thoracic cord were produced in 12 anesthetized, adult guinea pigs, and the outcome monitored, using successive behavioral tests and morphometry of the lesion at 2-3 months. Chronic histopathology was examined quantitatively with line-sampling of axons in 1-micron plastic sections through the lesion center, stained with toluidine blue. The type and distribution of damage to axons was similar to that seen following weight-drop contusion trauma in cats. Spinal cord function was examined by means of hindlimb reflex testing and motor behavior, vestibulospinal reflex testing, and mapping the receptive field of the cutaneus trunci muscle (CTM) reflex. These injuries characteristically resulted in a delayed onset of functional deficits at 1-2 days after injury, followed by partial recovery over the course of several weeks. Overall, functional outcome correlated significantly with the number of surviving axons in the lesion. The phenomenon of "secondary" pathology was striking at the behavioral level, whereas evidence of delayed injury has been indirect in most animal models. The onset of this secondary process occurred with a longer delay than has been assumed or implied by most suggested mechanisms of secondary pathology. The time course of secondary loss and recovery may be related to that of the inflammatory response at the injury site, particularly the phagocytic activity of macrophages.