We report U–Th disequilibria data for a suite of 13 young basaltic samples from the Samoan Islands, which represent the end-member mantle component EM2, and 4 historic lavas from Mt. Erebus, typifying young HIMU. The Samoan samples have low 230Th/232Th and 238U/232Th, consistent with the enriched nature (EM2) of the Samoan mantle source, whereas the Mt. Erebus samples have high 206Pb/204Pb and intermediate (230Th/232Th) and (238U/232Th). When considered in the context of the global oceanic basalt database, the Samoan samples' low 230Th/232Th and 238U/232Th greatly extend the global correlations between 230Th/232Th and 238U/232Th with 87Sr/86Sr and 143Nd/144Nd and change the functional form of these correlations from linear to hyperbolic. Using a maximum likelihood non-linear inversion method, we show that these correlations of 230Th/232Th and 238U/232Th with 87Sr/86Sr and 143Nd/144Nd can be approximated by two-component mixing. However, the global oceanic basalt data also show considerable scatter about the best-fit mixing curves. This scatter is attributed to additional source components and melting processes influencing the lavas (230Th/232Th) and (238U/232Th). Additional source components are supported by the Pb isotope data, which clearly require more than two endmember mantle components. We also show that the extent and variability in (230Th/238U) decreases as a function of source enrichment, with “depleted” MORB showing the largest extents and greatest variability in (230Th/238U) and “enriched” OIB, like Samoa, showing smaller extents and less variability of (230Th/238U). We interpret this observation in terms of differences in the melting regimes beneath mid-ocean ridges and ocean islands.