Experimental modeling of spinal cord injury is based mostly on mechanical effects, such as the impact of a weight dropped on the exposed spinal cord. The development of the resultant lesion is influenced by many interactive factors, e.g., the efficiencies of momentum and energy transfer to the cord and their profiles in time, and consequently the histopathologic reproducibility of the lesion is often inconsistent. We describe here a recoilless method that avoids these complications (as well as laminectomy) inherently. The vascular endothelium is injured photochemically, yielding chiefly small-vessel thrombosis and associated vasogenic edema sufficient to generate spinal cord necrosis to predetermined, reproducible degrees. This model is thus intended to simulate the secondary response of the vasculature to mechanical injury In the absence of hemorrhage. The most efficient version of this technique utilizes argon-dye laser excitation of the photosensitizing dye rose bengal at its 562-nm absorption maximum in tissue. With the laser beam focused in the shape of a thin (0.3-mm) line transverse to the spinal column at T8, a narrow zone of necrosis is initially produced. Within 1 week this initial zone expands in volume (length, 6-7 mm) to create a space that, when cleared of cellular debris, is suitable for cell Implantation. In cross section, striking features are the sharp horizontal demarcation between necrosed and viable tissue and the uniform progression of lesion depth as a function of Irradiation time. The necrotic region is bordered dorsally and laterally by a thin rim of viable tissue except at the beam focus; starting at 5 days there is evidence of demyelination in this peripheral region. By 14 days, myelination by oligodendrocytes and Schwann cells begins near this rim; large numbers of Schwann cells enter the dorsal cord at the epicenter, and myelinated axons occupy previously degenerated areas. By 2 months, the initial necrotic area begins to diminish, flattening laterally into clefts. Large, empty cavities develop secondarily with luminal surfaces that are smoothly contoured, as seen by electron microscopy, in contrast to the relatively irregular border of the initial necrotic lesion. Many of these morphological attributes mirror those observed after contusion injury. We anticipate that the uniformity and reproducibility of the photochemical lesion will prove useful in conducting statistically efficient tests of strategies, such as the administration of drugs, trophic factors, or transplanted cells, which are proposed to improve outcome after spinal cord injury.