A digital model study has been developed for quantitative assessment of experimental errors in the analysis of 133Xe clearance curve from the brain. A small computer synthesized a model of the clearance curve, varying combinations of fast and slow components. The curves were convoluted with Poisson random digits to simulate statistical fluctuations. Identical curves were overlapped with varying intervals to study the influence of remaining activity. The height over area method to ten minutes was confirmed to overestimate CBF by 10% to 15% with a slow component of 20 ml/100 gm per minute, and the overstimation was increased with a lower slow flow component. The initial slope value was shown to have a close relationship with the fast flow component when the latter was less than 100 ml/100 gm per minute. Errors due to statistical fluctuations were determined only by the initial height (Ho cps), as the percent standard deviation was deltaHo/Ho in the height over area method and 2 deltaHo/HologHo in the initial slope method, where deltaHo equals square rootHo. Remaining activity caused errors of 1% to 3% in the initial slope method with an injection interval of 15 minutes. The influence of remaining activity can be eliminated with an injection interval of more than 25 to 30 minutes in the initial slope method and more than 40 minutes in the height over area method.