The growth-associated protein-43 (GAP-43) is an axonal phosphoprotein which is expressed at high levels during development and is reinduced by regeneration in the PNS. Consequently it is believed to be a key molecule in the regulation of axonal growth. However, injury to the CNS does not result in significant regeneration and this has been suggested to correlate with a failure of central neurons to up-regulate GAP-43 after axotomy. We have examined a model of spinal cord injury which is unique in two respects; first dural integrity is maintained by compression of the cord with smooth forceps (thus excluding connective tissue elements) and, secondly, considerable axonal growth has been reported through the resulting lesion. Our previous studies have shown that GAP-43 is extensively distributed in the rat spinal cord (see accompanying paper), but here we have used anti-GAP-43 antiserum at a dilution which did not yield any immunostaining in normal cord. However, supranormal levels of GAP-43 were detected in cell bodies and axons around the lesion within four days of compression injury. Double immunostaining with the RT97 monoclonal antibody indicated that a small subpopulation of neurons local to the site of compression were axotomized and expressed GAP-43 and phosphorylated neurofilament epitopes in their cell bodies. Although damage to long axon tracts was extensive, there was no evidence of regeneration in white matter. On the other hand cavities which formed in grey matter provided an environment for axonal elongation. Immunolabelling with markers for astrocytes and endothelial cells was used to evaluate the interaction of elongating axons with endogenous CNS cell types. Sprouting axons, identified by the presence of elevated levels of GAP-43, did not appear to grow in contact with astrocytes but preliminary evidence suggested that newly formed capillaries provided an appropriate substrate.