Electron microprobe analysis (EMPA) of geological materials is often carried out assuming stoichiometry for one unanalyzed element to calculate matrix correction (ZAF) factors. Stoichiometric ZAF routines for multielement groups such as carbonate () are not commonly available. Consequently, carbonates are commonly analyzed by means of the O stoichiometry techniques applied to silicate and oxide minerals. Using real and simulated analytical data, we show that errors of up to 20% occur if ZAF corrections are made with the assumption of metal valence and stoichiometric O. Two techniques that allow the input of complete compositional data from analyses to ZAF factor calculations and that give accurate results are discussed. First, a new technique uses carbonate standards with compositions represented by the metal oxide MO4. The results are then recalculated as metal carbonate. The second method assigns incorrect valences to the elements, C is assumed to be stoichiometric, and O is anallzed directly. We show that an analytical total of 100% ± , where represents errors due to X-ray generation statistics, provides a powerful indication of analysis quality. The use of normalized and difference analysis methods, which necessarily assume totals of 100%, are to be avoided for this reason.