MCKAY and MITCHELL (1988) recently reported Ga con-centrations in whole rocks and mineral separates from gamet-and spinel-bearing peridotite xenoliths. They used these data to estimate the average Ga content of spine1 and garnet lher-zolites and suggested that an upper mantle, consisting of a 30:70% mixture of these respective components, would have an average Ga content of I .90 ppm with a minimum of 1.20 and a maximum of 2.58 ppm. They used this upper mantle estimate as a guide to the Ga abundance in the primitive mantle and from this perspective compared their estimate with the primitive mantle estimates of JAGOUTZ et al. (1979) and SUN (1982). They also discussed the relative distribution of Ga between the core and mantle. In view of the facts that (1) Ga is moderately incompatible during melting and (2) the peridotites studied by McKay and Mitchell are refractory, it is necessary to allow for the effects of melt extraction before the data can be used to discuss the issues relevant to the primitive mantle. By neglecting this procedure, the authors have reached some erroneous conclusions. McKay and Mitchell's estimated mantle Ga content of 1.9 ppm is much too low; when the peridotite data are corrected for the effects of melt extraction, a figure of 3.9 ppm Ga is obtained. This value agrees with that of JAGOUTZ et al. (1979) and SUN (1982). A fundamental question is, "How representative of the upper mantle are the McKay and Mitchell samples?" Their low A1203 and SC contents indicate that these peridotites, especially the garnet-bearing ones, are depleted relative to a pyrolite and primitive mantle. It is therefore necessary to compare their compositions with those of other peridotites, to determine how well they represent the present upper man-tle and how depleted they are relative to the primitive mantle. If peridotites are to be used to determine a primitive mantle abundance of Ga, what happened to them since accretion and core formation must be established. Most xenolith sam-ples, including those studied by McKay and Mitchell, have lost a melt component. This is reflected in their low A&O3 and SC contents and high Mg-value and Ni abundances rel-ative to the primitive mantle (JAGOUTZ et al., 1979; SUN, ZINDLER, 1986). To determine a primitive mantle Ga abundance the major and trace element (including Ga) concentrations must be modified to compensate for this fractionation. Ga is a mod-erately incompatible trace element and enriched in basalts relative to their source. Consequently, melt extraction will cause Ga depletion in the residual peridotite. To compensate for melt extraction it must be established how much basalt needs to be added to these peridotites to bring their refractory lithophile elements (e.g., Al, SC) in line with primitive mantle abundances. Only then can their Ga contents be modelled according to correlations between Ga and refractory lithophile elements. An additional constraint is that the garnet peridotite sam-ples of McKay and Mitchell, with their lower Al203 contents, are more depleted than spinel-bearing samples and thus re-quire the addition of more melt to compensate for this de-pletion. This means that more Ga would have to be added to the garnet peridotite samples than to the spinel-bearing samples, and this may indicate that there is little difference in the initial Ga contents between these peridotite types. A procedure to correct for melt extraction entails relating Ga to a major element, such as Al or a trace element such as SC. The relative variation of Alz03 and SC versus Ga in peridotite samples can be seen in Fig. 1: Ga correlates closely with A1203, whereas Ga versus SC shows some scatter. The wide range of Ga, SC, and A1203 contents in peridotite samples is attributed to the loss of partial melt. Independent estimates for the AlzOJ content of the prim-itive mantle are around 4.5% (JAGOUTZ et al., 1979; SUN, 1982; W~~NKE et al., 1984; PALME and NICKEL, 1985; HART and ZINDLER, 1986; MCDONOUGH and SUN, in prep.). Using this value and the A1203-Ga trend in Fig. la it is concluded that the primitive mantle has about 3.9 to 4.0 ppm Ga. This value is much higher than McKay and Mitchell's suggested mantle abundance of 1.9 ppm, but is consistent with the estimates of JAGOUTZ et al. (1979), SUN (1982), and W~~NKE et al. (1984) for a primitive mantle based on lherzolite xeno-liths (3.8 ppm) and Archean komatiites (4.5 ppm). MCKAY and MITCHELL (1988) have created chaos in their use of the terms "mantle," "upper mantle," and "pyrolite" (p. 2869). They use the term pyrolite as a separate entity and it is not clear how they relate it to mantle and upper mantle. They do not state which of their Ga estimates represents an