Calculations were made with ef ciency curves developed for the micro-ori ce impactor (MOI) to estimate errors in mass collected on individual stages due to uctuations in the ow rate during sampling of submicrometer particles. The sizes of these errors depend on the size distribution of the sampled aerosol and the level of ow rate uctuation. For a log-normally distributed particle population, mass errors due to ow rate uctuations were bimodally distributed about stages with cutpoints near the aerosol Mass Median Aerodynamic Diameter (MMAD). The largest errors occurred uniformly on the stage with the smallest cutpoint (here, 0.059 ¹m). These errors were asymmetric with respect to sign, which leads to a net error for a randomly uctuating ow rate. In general, mass errors increased with decreasing geometric standard deviation (¾ g ) and were substantially greater for populations with 0.5 ¹m MMADs than for those with 0.2 ¹m MMADs. The largest net errors for the former were 4, 110, and 560% for ¾ g of 1.2 and ow rate uctuations of §1, §5, and §10%, respectively, but decreased to 0.03, 0.9, and 4%, respectively, for a ¾ g of 1.9. Flow rate uctuations, therefore, lead to a positive bias in the geometric standard deviation inferred from the measured masses and reduce the user's ability to interpret differences in size distributions. To minimize these effects, we developed and tested a system for controlling the volumetric sampling rate through a MOI at 30 LPM with a precision of 0.06% (600 ms averaging; 0.67% for 5 ms averaging), a level of precision that allows for accurate relative calibration between ow systems and for which errors from ow rate uctuations are reduced to <1%, even for a very narrow aerosol (¾ g 1.2). Mass errors for an uncontrolled eld test were as large as ¡60%, but these were reduced to <0.22% in a comparable controlled eld test. In two replicate tests of the system, agreements between stage masses collected on MOI stage 7 (D 50 = 0.173 ¹m) of two simultaneously operated ow-controlled impactors sampling 0.2 ¹m diameter monodisperse test particles were 0.997 and 0.996, although differences as large as 4% were observed for some stages. The system is suitable for use with standard "Federal Reference Method" samplers.