We performed a series of synthesis experiments at 1 atm pressure to investigate the substitution mechanisms of 1+ and 3+ ions into olivine. Forsterite crystals were grown from bulk compositions that contained the element of interest (e.g. Li) and different amounts of additional single trace elements. By working at constant (major element) liquid composition and temperature we eliminated all compositional effects other than those due to the trace elements. Mineral-melt pairs were then analysed to determine the compositional-dependence of the partition coefficient (D), which corresponds to frac([element]mineral, [element]glass), and where [element] refers to weight concentration of the element in the respective phase. We find that Li forms a stable coupled substitution with Sc and, at above ∼500 ppm Sc in the crystal, Li+ and Sc3+ ions form an ordered neutral complex ([LiSc]). This complex dissociates at lower trace element concentrations and a second, concentration-independent, mechanism begins to dominate. This second solution mechanism is most likely 2Li+ ⇔ Mg2+ where one of the Li atoms is in an interstitial position in the crystal lattice. Natural olivines show Li contents slightly greater than Sc (on an atomic basis), indicating that both substitution mechanisms are significant. Unlike Sc, Al does not appear to form a stable complex with Li in the olivine structure. Sodium is highly incompatible in olivine with DNaFo s(-) melt of ∼0.00015-0.03. Olivine-liquid partitioning of Na+ is independent of Sc3+ or Al3+ concentration. This indicates that the coupled substitution of Na+ with any 3+ ions is unlikely. Instead, the relevant substitution mechanism appears to be 2Na+ ⇔ Mg2+. Although independent of 3+ ion concentration, DNaFo s(-) melt is inversely correlated with the Li concentration of both melts and crystals, implying that Na competes (unsuccessfully) with Li to replace Mg in the olivine structure. Aluminium is highly incompatible in forsterite fenced(DAlFo s(-) melt = 0.006 ± 0.0005 n = 7). Values DAlFo s(-) melt of are similar for all phase pairs synthesised from starting materials containing between 10 and 100,000 ppm Al. This suggests that Al is principally incorporated in forsterite by replacing one Mg and one Si atom fenced(MgMgx + SiSix ↔ AlMg• + AlSi′), where the Al atoms on octahedral (Mg) and tetrahedral (Si) sites are dissociated from one another. The incorporation of gallium into forsterite is influenced by the presence of Li. Where Li concentration in the crystal is much greater than that of Ga (on an atomic basis) we find an excellent correlation between DGaFo s(-) melt and melt Li content. This relationship indicates that Ga3+ and Li+ replace 2Mg2+ on octahedral sites and that the Ga and Li atoms are, like Sc and Li, strongly associated in the crystal structure. The mechanism by which scandium is incorporated into forsterite is strongly governed by the presence Li. As discussed above, ordered fenced(LiMg′ ScMg•) complexes form readily in forsterite in Li-rich experiments. Under Li-absent but Sc-rich conditions (Sc in the crystal >∼500 ppm), DScFo s(-) melt is proportional to the concentration of Sc in the melt. This indicates that Sc incorporation is charge-balanced by the formation of magnesium vacancies fenced(3 MgMgx ↔ 2 ScMg• + VMg″), and that both species are associated fenced(fenced(ScMg• VMg″)). At lower Sc concentrations (<500 ppm in the crystal), the concentration-dependence of partitioning indicates that the fenced(ScMg• VMg″) complexes dissociate. Our results demonstrate that partitioning of 1+ and 3+ ions into olivine is complex and involves a range of point defects which yield strongly composition-dependent crystal-melt partition coefficients. Since physical and chemical properties of natural olivine, such as diffusion of 6Li and 7Li and H2O solubility, depend on the concentrations of the defects identified in this study, our results provide an important insight into how determining substitution mechanisms can improve our understanding of large-scale mantle processes and properties.