New analyses of He, Ne, Ar and CO_2 trapped in basaltic glasses from the Southeast Indian Ridge (Amsterdam–St. Paul (ASP) region) show that ridge magmas degas by a Rayleigh distillation process. As a result, the absolute and relative noble gas abundances are highly fractionated with ^4He/^(40)Ar^* ratios as high as 620 compared to a production ratio of ∼3 (where ^(40)Ar^* is ^(40)Ar corrected for atmospheric contamination). There is a good correlation between ^4He/^(40)Ar^* and the MgO content of the basalt, suggesting that the amount of gas lost from a particular magma is related to the degree of crystallization. Fractional crystallization forces oversaturation of CO_2 because CO_2 is an incompatible element. Therefore, crystallization will increase the fraction of gas lost from the magma. The He–Ar–CO_2–MgO–TiO_2 compositions of the ASP basalts are modeled as a combined fractional crystallization–fractional degassing process using experimentally determined noble gas and CO_2 solubilities and partition coefficients at reasonable magmatic pressures (2–4 kbar). The combined fractional crystallization–degassing model reproduces the basalt compositions well, although it is not possible to rule out depth of eruption as a potential additional control on the extent of degassing. The extent of degassing determines the relative noble gas abundances (^4He/^(40)Ar^*) and the ^(40)Ar^*/CO_2 ratio but it cannot account for large (>factor 50) variations in He/CO_2, due to the similar solubilities of He and CO_2 in basaltic magmas. Instead, variations in CO_2/^3He (≡C/^3He) trapped in the vesicles must reflect similar variations in the primary magma. The controls on C/^3He in mid-ocean ridge basalts (MORBs) are not known. There are no obvious correlated variations between C/^3He and tracers of mantle heterogeneity (^3He/^4He, K/Ti etc.), implying that the variations in C/^3He are not likely to be a feature of the mantle source to these basalts. Mixing between MORB-like sources and more enriched, high ^3He/^4He sources occurs on and near the ASP plateau, resulting in variable ^3He/^4He and K/Ti compositions (and many other tracers). Using ^4He/^(40)Ar^* to track degassing, we demonstrate that mixing systematics involving He isotopes are determined in large part by the extent of degassing. Relatively undegassed lavas (with low ^4He/^(40)Ar^*) are characterized by steep ^3He/^4He–K/Ti mixing curves, with high He/Ti ratios in the enriched magma (relative to He/Ti in the MORB magma). Degassed samples (high ^4He/^(40)Ar^*) on the other hand have roughly equal He/Ti ratios in both end-members, resulting in linear mixing trajectories involving He isotopes. Some degassing of ASP magmas must occur at depth, prior to magma mixing. As a result of degassing prior to mixing, mixing systematics of oceanic basalts that involve noble gas–lithophile pairs (e.g. ^3He/^4He vs. ^(87)Sr/^(86)Sr or ^(40)Ar/^(36)Ar vs. ^(206)Pb/^(204)Pb) are unlikely to reflect the noble gas composition of the mantle source to the basalts. Instead, the mixing curve will reflect the extent of gas loss from the magmas, which is in turn buffered by the pressure of combined crystallization–degassing and the initial CO_2 content.