A negative relation exists between the steady-state, cerebrospinal fluid/blood concentration ratio and the hydrodynamic radius of blood-derived proteins. This negative relation can be explained if blood proteins enter spinal fluid only at the choroid plexus (and not at other sites within the brain), and if transfer at the plexus is limited by the choroidal epithelium and not by demonstrably leaky capillaries within the choroidal stroma. The most likely interpretation of the data is that there are two pathways for protein transfer at the plexus epithelium, a set of 117-Å-radius pores which allow transfer of smaller proteins by diffusion, ultrafiltration, or both mechanisms, and a set of 250-Å-radius pinocytotic vesicles which account for exchange of larger proteins as well. On the basis of the data, it is possible to calculate permeability and/or osmotic reflection coefficients associated with porous transfer, as well as a rate of vesicular transport. The latter is estimated to be so low that ultrastructural studies would be expected to demonstrate only occasional vesicular discharge from the choroidal epithelium into cerebrospinal fluid. Pores with a 117-Å-radius have not been described at the choroidal epithelium, but may represent a 0.08% defect in the normally continuous tight junctions that surround and closely connect choroidal epithelial cells. © 1979.