Computer simulations and error analysis of a simple two-compartment, passive-diffusion exchange model for 82Rb across the blood-brain-barrier (BBB) have demonstrated the feasibility of obtaining useful estimates of unidirectional rate constant (K1) and tissue-blood water volume (Vb) in vivo using dynamic positron tomography (PET). The coefficients of variation (CV) in parameter estimates for 20 studies on ten patients are shown to have mean values of 10% for tumour K1 and Vb, 6% for normal brain Vb and 30% for normal brain K1. Ten test-retest studies show a very high correlation (R2 greater than 0.88) between the estimated parameter values. Apparent tissue blood volume (Vb) estimates obtained from 82Rb studies underestimate the blood volumes obtained by single-breath C15O studies by approximately 15%. In normal brain the extraction of rubidium (E) is small (less than 10%), and gives rise to a linear relationship between K1 and permeability-surface area product (PS). In tumour, however, E is larger (greater than 30%), and results in a non-linear relationship between these values. For both tumour and normal brain, K1 was found to be independent of regional cerebral blood flow (rCBF). The data from these clinical studies are in agreement with the predictions of the computer simulations and suggest the suitability of 82Rb/PET in the quantification of BBB functional changes.