Accurate determination of the reproducibility of measurements using the microsphere technique is important in assessing differences in blood flow to different organs or regions within organs, as well as changes in perfusion under various experimental conditions. The sources of error of the technique are briefly reviewed. In addition, we derived a method for combining quantifiable sources of error into a single estimate that was evaluated experimentally by simultaneously injecting eight or nine sets of microspheres (each with a different radionuclide label) into four anesthetized dogs. Each nuclide was used to calculate blood flow in 145-190 myocardial regions. We compared each flow determination (using a single nuclide label) with a weighted mean for the piece (based on the remaining nuclides). The difference was defined as "measured" error. In all, there were a total of 5,975 flow observations. We compared measured error with theoretical estimates based on the Poisson error of radioactive disintegration and microsphere entrapment, nuclide separation error, and reference flow error. We found that combined estimates based on these sources completely accounted for measured error in the relative distribution of microspheres. In addition, our estimates of the error in measuring absolute flows (which were established using microsphere reference samples) slightly, but significantly, underestimated measured error in absolute flow.