It has been more than two decades since White (1985) and Zindler and Hart (1986) proposed that the observed range of Sr-Nd-Pb isotope ratios of oceanic basalts can be described as mixtures of depleted mantle (DMM) with a limited number of enriched global endmember components (HIMU, EMI, EMII). There is no doubt that the global endmembers in isotope space represent extremes of the timing and magnitude of chemical fractionation processes in the Earths mantle. However, it remains a matter of debate how the intermediate isotopic compositions often evident on the local scale of individual islands are formed: (1) Do they represent mixtures between the limited number of global endmembers, or (2) do they reflect processes intermediate to the ones forming the global endmembers, or (3) does each individual ocean island basalt suite provide us with information about the timing and kind of geochemical differentiation forming that single source? Each possibility has important geodynamic implications. The Earths mantle is continuously differentiating through partial melting and remixing through plate tectonic recycling and convection, suggesting variable timing and composition of mantle sources. However, mantle sources might be formed by mixed lithologies and thus melt compositions might reflect mixtures of sources of more extreme compositions, possibly representing the global endmembers. We further address these questions based on two examples. Grande Comore Island is located on 140 Ma Indian Ocean lithosphere and its lavas reflect plume-lithosphere interaction. The Grande Comore plume component has Sr-Nd-Pb isotopic compositions intermediate between HIMU and EMI. Its extreme Os isotope ratios are among the highest measured in shield building-stage lavas of oceanic islands, giving further support for generally radiogenic Os isotope ratios in the EMI and HIMU compositions. A lack of correlation between OIB with high Os isotope ratios with inferred lithospheric thickness implies that they are not solely controlled by melt dynamics of a pyroxenite-peridotite source, but require variable proportions of pyroxenite in individual sources. New isotope data from the second example, the Discovery Seamounts in the South Atlantic, reveal a continuum in compositions between the extreme EMI composition of Walvis Ridge DSDP 525A and the LOMU extreme of the Discovery ridge anomaly (Douglass et al., 1999) and require a range of extreme composition outside the mixing tetrahedron of the global endmembers. In the global context, each individual island or volcano with enriched mantle affinity seems to form a trend towards its own unique enriched mantle endmember, inconsistent with mixing between narrowly defined global endmembers. The spectrum of enriched mantle endmembers is consistent with a dynamic Earth, continuously recycling varying proportions of oceanic crust, sediment and some continental lower crust or mantle. Douglass, J., Schilling, J.-G. and Fontignie, D., 1999. J. Geophys. Res., 104: 2941-2962.White, W.M., 1985. Geology, 13: 115-118.Zindler, A. and Hart, S., 1986. Ann. Rev. Earth Planet. Sci., 14: 493-571.