Routes of vascular leakage resulting in trauma-induced edema have not been clarified. To explore the problem we followed the fate of intravascular horseradish peroxidase (HRP) after compression injury to the thoracic cord (cats). At 90 seconds and 15 minutes, HRP was confined to the gray matter, occupying perivascular spaces, unexpanded extracellular channels, and cytoplasmic compartments of injured cells. Adjoining segments contained similar but lesser deposits. At 4 hours, tracer occupied the expanded extracellular spaces of the lesion's white matter; gray matter deposits were present up to 4 cm. distal. Vessels revealed no evidence of rupture. Open interendothelial junctions were not found. Counts of HRP-labeled vesicles in the endothelium of gray matter capillaries revealed a significant intensification of vesicular activity in the lesion and in adjacent areas up to 9 cm. caudal. Morphologically, labeled vesicles exhibited a wide diversity in shape and size. Typical pinocytotic (700A) and tubular forms measured 400 to 700 A in width; vacuolar forms measuring up to 0.7 mum. across were frequently observed. Continuity between the three types was often evident. Where basement membrane and perivascular clefts were not yet inundated with HRP, sites of vesicular emptying of HRP at the tissue front were identified. Serial sections revealed that vesicles may be contiguous from luminal to abluminal surfaces, thus providing facilitated transport pathways. The data suggest that vesicular transport plays a role in the genesis of trauma-induced edema.