The pattern of axonal destruction and demyelination that occurs in experimental contusion injury of cat thoracic spinal cord was studied by line sampling of axons in 1 micron thick plastic sections with the light microscope. Injuries were produced by a weight-drop apparatus, with the vertebral body (T9) below the impact stabilized by supports under the transverse processes. The effects of two combinations of weight and height were examined: 10 or 13 g dropped 20 cm onto an impact area of 5 mm diameter. Animals were kept for 3-5 months after injury, then fixed by perfusion for histological analysis. The number of surviving myelinated axons was found to vary both with the weight used and with the size of the spinal cord. A measure of impact intensity was derived from the calculated momentum of the weight at impact divided by the cross sectional area of the cord (interpolated from dimensions measured rostral and caudal of the lesion following fixation). At impact intensities greater than 0.02 kg-m/s/cm2 there was practically no survival of axons at the center of the injury site, combined with almost complete breakdown of the pial margin. Between 0.08 and 0.2 kg-m/s/cm2 the number of surviving axons varied between 100,000 and 2,000, approximating a negative exponential function (r = -0.88). The number of axons surviving in the outer 100 microns of the cord varied practically linearly (r = -0.82) between near normal and less than 1% of normal over the same range of injury intensity. The number of surviving axons decreased with depth from the pia, also approximating a negative exponential function, with a 10-fold decrease in density over approximately 500 microns. The average slope of this relation with depth remained similar over the range of injury intensity examined, though the slope appeared inversely related to variation in axonal survival for different individuals at a given intensity. It is argued that the loss of axons is probably determined primarily by mechanical stretch at the time of impact. Its centrifugal pattern may be explained by longitudinal displacement of the central contents of the cord, reflecting the viscoelastic "boundary layer" properties of parenchymal flow within the meningeal tube. This is illustrated with reference to the behavior of a gelatin model under compression. The preferential loss of large caliber axons and the characteristic shift to abnormally thin myelin sheaths (resulting from post-traumatic demyelination) both varied in extent independently of injury intensity and overall axonal survival.(ABSTRACT TRUNCATED AT 400 WORDS)