INTERACTION of melts ascending from the asthenosphere with the lithospheric mantle through which they pass is widely accepted to be a significant process by which the lithosphere can acquire distinctive trace-element and isotopic signatures(1-3). Despite the importance of this process in the genesis of the lithosphere, the nature of these ’metasomatic’ reactions is still debated(2-5). Here I report the results of an experiment designed to investigate the mechanism by which such metasomatism could occur, A natural sample of Group I kimberlite, a low-volume partial melt of asthenospheric origin, is reacted with a refractory peridotite (harzburgite) in a thermal gradient at high pressure to simulate the interaction between a dyke and its host rock in the lower lithosphere. The harzburgite develops several contrasting mineralogical zones, the coolest of which contains olivine, orthopyroxene, clinopyroxene, phlogopite, amphibole, Nb-Cr-rutile and minor zircon. The formation of mica and amphibole is accompanied by increases in trace-element concentrations and Rb/Sr, Rb/Ba and Pb/U ratios-features characteristic of metasomatized xenoliths recovered from southern African kimberlites(5). These results support the suggestion, based on temporal association and isotopic data, that at least some mantle metasomatism is related to the interaction of Group I kimberlite with refractory peridotite in the lower lithosphere(6).