Based on composition and abundance of garnets from heavy mineral concentrates and from ultramafic nodules, it is estimated that low-calcium garnet harzburgites, thought to be the host rocks of most diamonds in the upper mantle, form 2% of the mantle (5% of the garnet-bearing portion) sampled by Group I kimberlites at Kimberley, South Africa. As both sampling approaches (xenocryst abundance and xenolith abundance) yield virtually identical estimates, there is no need to invoke the presence of magnesite or interstitial liquid to cause disaggregation of low-Ca garnet harzburgites during eruption. Six of the eleven low-Ca garnet harzburgite xenoliths identified contain zoned garnets and although there is considerable variation between samples, typically Ca increases and Cr decreases from core to rim. All other minerals are homogeneous (except in one rock) and magnesian (e.g., Mg/(Mg+Fe) in olivine=0.929-0.949). Four of the eleven samples contain magnesio-chromite (59.7--62.6 wt % Cr2O3), and two have primary phlogopite. These rocks have apparently equilibrated at pressure and temperature conditions equivalent to those on a conductive subcontinental geothermal gradient, mostly in the diamond stability field, and within the range of equilibration of the majority of garnet lherzolites from Kimberley. Zoning is interpreted as due to re-equilibration of the Ca-poor domains toward the local bulk composition of the diopside-saturated lherzolite upper mantle. Introduction of Ca may have been fluxed by the presence of a fluid (also introducing phlogopite), but uniformly low TiO2 values argue against interaction with a silicate melt. Such Ca metasomatism, if post diamond formation, could ultimately result in transformation of low-Ca garnet harzburgites into garnet lherzolites that could contain diamonds with low-Ca garnet harzburgite paragenesis inclusions. ¿ American Geophysical Union 1995