Recently, a new model of the urinary concentrating process has been proposed that takes into account the three-dimensional architecture of the renal medulla. Under the assumptions of the model, computer simulations predicted significant axial osmolality gradients in the inner medulla without active transport by the inner medullary loop of Henle. Two of the model assumptions (which constitute hypotheses for this study) were: (1) the osmotic water permeability of the initial part of the inner medullary collecting duct (initial IMCD) is very low even in the presence of vasopressin; and (2) there is significant lateral separation of structures such that thin descending limbs are far from the collecting ducts at the same inner medullary level. The first hypothesis was addressed by perfusing rat initial IMCD segments in vitro and measuring osmotic water permeability. With the osmotic gradient oriented as predicted by the model (lumen greater than bath), vasopressin increased the osmotic water permeability from 286 to 852 microns/s. Three additional series of experiments confirmed the high water permeability in the presence of vasopressin. The second hypothesis was addressed by morphometric analysis of histologic cross-sections of the rat renal medulla. Mean distances of descending limbs to the nearest adjacent collecting duct were very small throughout the inner medulla (less than 6 microns) and substantially less than in the outer medulla (28 microns). It was concluded that the data are inconsistent with both hypotheses and therefore do not support the feasibility of the "three-dimensional" model of the renal inner medulla. The axial distributions of loops of Henle and collecting ducts in the rat renal medulla are also reported.