Garnets from diamondiferous and non-diamondiferous peridoties nodules from the Roberts Victor kimberlite were analysed for trace elements by secondary ion mass spectrometry (SIMS). Comparison with inclusions in diamonds from Akwatia (Ghana) show that there are no significant difference between garnets in peridotite xenoliths and those included in diamonds. High chromium contents (Cr2O3 generally >5 wt%) indicate that both harzburgitic and iherzolitic garnets from Roberts Victor (nodules) and Akwatia (diamonds) were crystallised from precursor rocks that had suffered depletion by partial melting in the spinel stability field which resulted in high Cr/Al in the residue. Harzburgitic and lherzolitic garnets, despite their common spinel harzburgite protoliths, show distinctive trace element chemistries that are not transitional to each other. This contrasts with the observation of concordant overlapping trends for a number of major elements and some trace elements for southern African garnet harzburgite and lherzolite xenoliths [F.R. Boyd, Earth Planet, Sci. Lett. 96 (1989) 15-26]. The REE patterns of harzburgitic garnets, when normalised to a 'primitive' garnet composition, are depleted in MREE and HREE and enriched in LREE. A positive slope within the HREE, particularly prominent in inclusions in diamonds, points to depletion (partial melting) originating in the garnet stability field. We suggest that the protlith had undergone polybaric fractional melting (as envisaged for modern MORB extraction) before being subducted to the depth of the diamond stability field. The growing garnets acquired a depleted REE pattern which was subsequently overprinted by a metsomatising fluid with very high LREE/HREE. The lherzolitic garnets, despite their high chronium contents, have flat REE patterns when normalised to a primitive garnet, a paradox which requires a secondary enrichment event to produce the unfractionated REE abundances. Lower LREE in lherzolitic relative to harzburgitic garnets may be attributed to re-equilibration with clinopyroxene. Differences in MREE and particularly in HREE cannot be explained by such a process, due to strongly increasing DGrt/Cpx from LREE to HREE. Therefore, lherzolitic garnets indicate an additional metasomatic process not seen by their harzburgitic counterparts, which introduced major elements (such as Ca, Fe and Si) sufficiently to stabilise clinopyroxene and which re-enriched the previously depleted MREE and HREE.