Introduction: Both the chemical and isotopic bulk compositions of Types A (melilite-rich) and B (pyroxene-rich) refractory inclusions from CV3 chondrites suggest that they may be evaporative residues of high-temperature condensate assemblages [1]. The MgO and SiO2 contents of these objects are generally consistent with the degrees of evaporation implied by the isotopic compositions that have been measured, but direct evidence of this process is sparse because there are few inclusions for which bulk chemical and bulk Mg and Si isotopic data exist. We are addressing this problem by obtaining these data for a suite of inclusions from Allende, Efremovka, and Leoville. Here, we report new bulk Mg isotopic compositions for inclusions whose bulk chemical compositions we have recently determined [2]. Their bulk Si and O isotopic compositions will also be measured as part of this study, allowing us to calculate their original bulk compositions based on the implied amounts of evaporative losses. Sampling and analytical techniques: For each inclusion, 1 or 2 thin sections were prepared. Material was excavated from the remaining portions, ground into a homogeneous powder (5-200 mg, most ~40 mg) and divided into aliquots for instrumental neutron activation analysis (INAA), bulk Mg isotopic analysis, and bulk Si and O isotopic analysis. Our modal recombination and INAA methods are described elsewhere [2]. Bulk Mg isotopic data were obtained with the IsoProbe multi-collector, magnetic sector ICP-MS at LLNL and the Nu-Plasma double focussing MC-ICP-MS at U. Cambridge. For Mg isotopic analysis, powders were dissolved in acid and insoluble residues were removed. Mg was purified on resin columns with quantitative recovery. Analytical techniques are similar to those described in [3]. Results: The Mg isotopic data are given in Table 1 and are reported relative to the DSM3 standard [4]. Analyses of 6 of 7 inclusions analyzed by both labs are within analytical uncertainty of each other, demonstrating there is no significant intra-laboratory bias between LLNL and UC. The results from the two labs were averaged together for each of these samples, and the errors were mathematically propagated. For the other sample, TS32, the uncertainty given below represents the range of the two analyses. There is no clear relationship between isotopic composition and petrographic type. The inclusion with the highest FMg is TS32, a compact Type A (CTA), but other Table 1. Mg Isotope Compositions of CAIs.