This study presents a chemical protocol for the separation of Mg that is particularly adapted to alkali-rich samples (granite, soil, plants). This protocol was based on a combination of two pre-existing methods: transition metals were first removed from the sample using an AG-MP1 anion-exchange resin, followed by the separation of alkalis (Na, K) and bivalent cations (Ca 2+ , Mn 2+ and Sr 2+) using a AG50W-X12 cation-exchange resin. This procedure allowed Mg recovery of ~100 ± 8%. The [Σcations]/[Mg] molar ratios in all of the final Mg fractions were lower than 0.05. The Mg isotope ratios of eleven reference materials were analysed using two different MC-ICP-MS instruments (Isoprobe and Nu Plasma). The long-term reproducibility, assessed by repeated measurements of Mg standard solutions and natural reference materials, was 0.14‰. The basalt (BE-N), limestone (Cal-S) and seawater (BCR-403) reference materials analysed in this study yielded δ 26 Mg mean values of-0.28 ± 0.08‰,-4.37 ± 0.11‰ and-0.89 ± 0.10‰ respectively, in agreement with published data. The two continental rocks analysed, diorite (DR-N) and granite (GA), yielded δ 26 Mg mean values of-0.50 ± 0.08‰ and-0.75 ± 0.14‰, respectively. The weathering products, soil (TILL-1) and river water (NIST SRM 1640), gave δ 26 Mg values of-0.40 ± 0.07‰ and-1.27 ± 0.14‰, respectively. We also present, for the first time, the Mg isotope composition of bulk plant and organic matter. Rye flour (BCR-381), sea lettuce (Ulva lactuva) (BCR-279), natural hairgrass (Deschampsia flexuosa) and lichen (BCR-482) reference materials gave δ 26 Mg values of-1.10 ± 0.14‰,-0.90 ± 0.19‰,-0.50 ± 0.22‰ and-1.15 ± 0.27‰ respectively. Plant δ 26 Mg values fell within the range defined by published data for chlorophylls.