We tested the reproducibility of consecutive measurements of regional cerebral blood flow (rCBF) made in 13 areas of each hemisphere in patients with cerebrovascular diseases by the atraumatic 133Xe intravenous injection method. The data were analysed by a two-compartment model similar to that used in the Obrist inhalation method. Four parameters derived from the model were tested: Fg (flow of the fast-clearing compartment), FF and W (respectively fractional flow and relative tissue weight of the same compartment), ISI (initial slope index as defined by Risberg et al.). No significant variation was found in these four parameters between two consecutive rest measurements for all the areas studied and whatever the time interval between the measurements. The variances of the differences of Fg and ISI between the two measurements were found to be similar to those found in normal healthy subjects. The variances of FF and, consequently, W were greater than the values determined in healthy volunteers. An attempt to improve the quality of the results, by correcting them for the effects of PaCO2 changes between the measurements, resulted on the contrary in a general increase of the variances of the differences, showing the inadequacy of the correction coefficients used for healthy subjects at least for a proportion of the patients with cerebrovascular diseases. Sensitivity, tested by activation (hand work), shows a positive answer for FG but no significant change in ISI in any area. These findings demonstrate that in cases of cerebrovascular diseases the raw data (uncorrected for the changes in PaCO2) are the most reliable data but the meaning of the values found for FF and W must be regarded with caution since they may be affected by the “slippage” phenomenon. It is suggested that the ISI should only be used when there is evidence of a failure of the compartmental model, but only as a rough estimation of the flow level because of its lack of sensitivity.