The spinel–garnet transition in Cr/Al-enriched peridotitic bulk compositions is known from experimental investigations to occur at 20–70 kbar, within the pressure range sampled by kimberlites. We show that the Cr₂O₃–CaO compositions of concentrate garnets from kimberlite have maximum Cr/Ca arrays characterized by Cr₂O₃/CaO ∼0·96–0·81, and interpret the arrays as primary evidence of chromite–garnet coexistence in Cr-rich harzburgitic or lherzolitic bulk compositions derived from depth within the lithosphere. Under Cr-saturated conditions on a known geotherm, each Cr/Ca array implicitly delineates an isobar inside a garnet Cr₂O₃–CaO diagram. This simplification invites a graphical approach to calibrate an empirical Cr/Ca-in-pyrope barometer. Carbonaceous chromite–garnet harzburgite xenoliths from the Roberts Victor kimberlite tightly bracket a graphite–diamond constraint (GDC) located at Cr₂O₃ = 0·94CaO + 5·0 (wt %), representing a pivotal calibration corresponding to 43 kbar on a 38 mW/m² conductive geotherm. Additional calibration points are established at 14, 17·4 and 59·1 kbar by judiciously projecting garnet compositions from simple-system experiments onto the same geotherm. The garnet Cr/Ca barometer is then simply formulated as follows (in wt %): A small correction to P₃₈ values, applicable for 35–48 mW/m² conductive geotherms, is derived empirically by requiring conventional thermobarometry results and garnet concentrate compositions to be consistent with the presence of diamonds in the Kyle Lake kimberlite and their absence in the Zero kimberlite. We discuss application of the P₃₈ barometer to estimate (1) real pressures in the special case where chromite–garnet coexistence is known, (2) minimum pressures in the general case where Cr saturation is unknown, and (3) the maximum depth of depleted lithospheres, particularly those underlying Archaean cratons. A comparison with the P_Cr barometer of Ryan et al. (1996, Journal of Geophysical Research 101, 5611–5625) shows agreement with P₃₈ at 55 ± 2 kbar, and 6–12% higher P_Cr values at lower P₃₈. Because the P_Cr formulation systematically overestimates the 43 kbar value of the GDC by 2–6 kbar, we conclude that the empirical Cr/Ca-in-garnet barometer is preferred for all situations where conductive geotherms intersect the graphite–diamond equilibrium.