The garnet-orthopyroxene Al barometer specifically considers the Al content of orthopyroxene in equilibrium with garnet resulting from Mg-Tschermaks substitution. It is demonstrated that P-T calibrations of this barometer derived solely from experimental data for the MAS system, such as that favoured by Finnerty and Boyd (1984, 1987) based on the data of MacGregor (1974), cannot be expected to yield meaningful pressure estimates for natural garnet lherzolite assemblages. The presence of additional CaO, FeO and Cr2O3 components in natural garnet lherzolites can be expected to influence substantially the Al partitioning between orthopyroxene, garnet and/or spinel at any particular P and T. Thus a more comprehensive barometer formulation is required, such as the one provided by Nickel and Green (1985) that is based on experimental data for the CMAS and SMACCR systems with thermodynamic modelling and addition of an Fe correction term.

It is further emphasised that for orthopyroxenes in natural garnet lherzolites the amount of Al introduced as Mg-Tschermaks substitution cannot be assessed simply as the total Al cation content since such orthopyroxenes frequently contain Al cations linked to Na substitution in M2 sites or to Cr, Ti and possibly Fe3+ in M1 sites. Revised algorithms for the calculation of specific orthopyroxene contents are presented. Application to analytical data sets for garnet lherzolite zenolith suites in the Thaba Putsoa and Mothae kimberlites generates revised upper mantle P-T arrays which refute the widely accepted advocacy by Finnerty and Boyd (1984, 1987) and Finnerty (1989) of an upper-mantle palaeogeotherm beneath northern Lesotho that is markedly inflected to a higher thermal gradient at the depths of derivation of the more chemically fertile, porphyroclastic textured, xenoliths.