BaSO4 precipitated from mixed salt solutions by common techniques for SO isotopic analysis may contain quantities of H2O and NO that introduce errors in O isotope measurements. Experiments with synthetic solutions indicate that δ18O values of CO produced by decomposition of precipitated BaSO4 in a carbon reactor may be either too low or too high, depending on the relative concentrations of SO and NO and the δ18O values of the H2O, NO, and SO. Typical δ18O errors are of the order of 0.5 to 1‰ in many sample types, and can be larger in samples containing atmospheric NO, which can cause similar errors in δ17O and Δ17O. These errors can be reduced by (1) ion chromatographic separation of SO from NO, (2) increasing the salinity of the solutions before precipitating BaSO4 to minimize incorporation of H2O, (3) heating BaSO4 under vacuum to remove H2O, (4) preparing isotopic reference materials as aqueous samples to mimic the conditions of the samples, and (5) adjusting measured δ18O values based on amounts and isotopic compositions of coexisting H2O and NO. These procedures are demonstrated for SO isotopic reference materials, synthetic solutions with isotopically known reagents, atmospheric deposition from Shenandoah National Park, Virginia, USA, and sulfate salt deposits from the Atacama Desert, Chile, and Mojave Desert, California, USA. These results have implications for the calibration and use of O isotope data in studies of SO sources and reaction mechanisms. Published in 2008 by John Wiley & Sons, Ltd.