Although fully explored in larger animals, the role of injection site and sample microsphere content on variability of coronary blood flow (CBF) measurement using the microsphere technique remains controversial in rats despite the fact that this species is extensively used in cardiovascular research. We therefore investigated these variables in two studies. In a first study, we established that the precision of the method, assessed by the variability of four simultaneous CBF determinations, was a function of the sample microsphere number. Coefficient of variation (CV) averaged 4-10% when the tissue and reference samples received greater than 1000 and greater than 100 spheres, respectively, and did not improve appreciably with larger numbers of microspheres. In a second study, flow CV was measured following left atrial (LA) or left ventricular (LV) microsphere injections performed nearly simultaneously in the same conscious animal or in two similar groups of animals. CBF variability was lower by 22-62% after LA than after LV injections. Estimates obtained from separate analysis of the main variability components indicated that, with one exception, the variability associated with LV injections was at least 1.4 to 2.8 times higher than that due to LA injections. These findings establish the minimum number of microspheres needed to obtain precise blood flow determinations in the rat model and confirm previous reports, in anaesthetised rats, that LA microsphere injections generally yield more precise coronary blood flow determinations than LV injections.