The water permeability of capillaries in the subfornical organ (SFO) of rat was measured by a ¹H nuclear magnetic resonance method in combination with a venous injection of a relaxation reagent, gadolinium-diethylene triamine-N,N,N'âN'',N''-pentaacetic acid (Gd-DTPA^2-), which could not pass through the blood-brain barrier (BBB). Judging from results of Gd-DTPA^2- dose dependency in the intact brain and the BBB-permeabilized brain, Gd-DTPA^2- could not have leaked out from the capillaries in the cortex, thalamus or SFO, but it could have been extravasated in the posterior lobe of the pituitary gland. The longitudinal (T₁) relaxation time of water in the SFO region was measured by inversion-recovery magnetic resonance imaging at 4.7 T. The T₁ relaxation rates (1/T₁) before and after Gd-DTPA^2- infusion were 0.70 ± 0.02 s^-1 (mean ± S.E.M., n = 9) and 1.53 ± 0.11 s^-1 (n = 9), respectively. The rate constant for water influx to the capillaries was estimated to be 0.84 ± 0.11 s^-1 (n = 9) which corresponds with a diffusive membrane permeability (P_d) of 3.7 x 10^-3 cm s^-1. Compared with values found in the literature available on this subject, this P_d value for the capillaries in the SFO was the same order of magnitude as that for transmembrane permeability of water for the vasa recta, and it may be 10-100 times larger than that of the blood-brain barrier in the cortex. Areas of the cortex and thalamus showed minimal changes in the T₁ relaxation rate (ca 0.09 s^-1), but these values were not statistically significant and they corresponded to P_d values much smaller than those found in the SFO. From these results, we conclude that the capillaries in the SFO have one of the highest water permeability values among all of the capillaries in the brain. It is also suggested that this magnetic resonance imaging, based on T₁ relaxation rate, is a useful method to detect local water permeability in situ.