The reliability and accuracy of the bolus injection-dye dilution technique were assessed for a physiological range of frequencies (13-49 min-1) and stroke volumes (60-160 ml) on a glass model where flows and volumes varied as a preset function of time (n = 320). We found that the technique overestimates flow by about 8% with a 95% confidence interval of +/- 10% for one measurement. Mean transit times are accurate within a +/- 7% confidence interval for one measurement. In a time-dependent flow and volume system this technique measures the mean volume as related to time with fluctuations up to +/- 30% around the mean. Results are independent of time and site of injection. The double injection-single sampling technique gives results that are equivalent to those obtained by single injection and sampling of dye.