There is an intriguing paper just out: Konter, J.G., Barry B. Hanan, Janne Blichert-Toft, Anthony A.P. Koppers, Terry Plank, Hubert Staudigel, One hundred million years of mantle geochemical history suggest the retiring of mantle plumes is premature, Earth Planet. Sci. Lett., in press, 2008. It provides confirmation of an interpretation provided earlier in the same journal: Pilger, R. H., 2008, Discussion of “Break-up spots: Could the Pacific open as a consequence of plate kinematics”, by Clouard and Gerbault, Earth Planet. Sci. Lett. (See also: "Hotspots" of the South Pacific).
That is, three melting anomalies in the South Pacific are not only responsible for 20 Ma and younger island/seamounts, but also much older seamounts from the Tokelau, Gilbert, and Magellan seamounts to the northwest. Add in the Foundation and Austral seamounts as another melting anomaly trace -- that makes four semi-aligned "hotspots" in the region: Foundation --Foundation and Austral seamounts; Macdonald - part of Cook Islands and Tokelau seamounts; Cook - part of Cook Islands and Gilbert seamounts; Samoa - Samoan islands and Magellan seamounts.
There are apparent gaps of tens of millions of years in activity of the melting anomalies. It is difficult to see how mantle plumes could explain persistent, but intermittent activity. Perhaps mantle heterogeneities, combined with lithospheric extension and/or plate thickness variations could explain the intermittency. (See also: GSA Article...The Bend).