The protein tracer, horseradish peroxidase (HRP), was infused into the lateral cerebral ventricles or subarachnoid space of anesthetized cats and dogs after insertion of a cisternal cannula to permit drainage of cerebrospinal fluid (CSF) and tracer solution. The intracerebral distribution of the tracer was then determined by light microscopy of serial brain sections after postinfusion intervals of 4 min-2 h. For the localization of HRP, sections were incubated with diaminobenzidine (DAB) or the much more sensitive chromogen, tetramethylbenzidine (TMB). The TMB reaction showed a consistent 'paravascular' distribution of tracer reaction product, within the perivascular spaces (PVS) around large penetrating vessels and in the basal laminae around capillaries, far beyond the termination of the PVS. After infusion of HRP over 4 min, arterioles were surrounded by the tracer, but capillaries and venules were usually less densely demarcated; by 6 min, however, the intraparenchymal microvasculature was outlined in toto throughout the forebrain and brainstem. Electron microscopy of sections incubated in DAB after 10 or 20 min HRP circulation confirmed the paravascular location of the reaction product, which was also dispersed throughout the extracellular spaces (ECS) of the adjacent parenchyma. Our results demonstrate that solutes in the CSF have access to the ECS throughout the neuraxis within minutes via fluid pathways paralleling the intraparenchymal vasculature. The rapid paravascular influx of HRP could be prevented by stopping or diminishing the pulsations of the cerebral arteries by aortic occlusion or by partial ligation of the brachiocephalic artery. The exchange of solutes between the CSF and the cerebral ECS has generally been attributed to diffusion, however, HRP enters the neuraxis along the intraparenchymal microvasculature far more rapidly than can be explained on this basis. This apparent convective tracer influx may be facilitated by transmission of the pulsations of the cerebral arteries to the microvasculature. We postulate that a fluid circulation through the CNS occurs via paravascular pathways.