An empirical calibration for the quartz-calcite thermometer was derived from measured DELTA O-18(qz-cc) values from greenschist-facies marbles, veins composed of cogenetic quartz and calcite, and various low-grade metamorphic rocks. The DELTA O-18(qc-cc) values vary systematically with independently determined formation temperature and can be fit to the expression 1000 In alpha(qz-cc) = 0.87 (+/- 0.06) x 10(6)/T2. In contrast, published results from direct-exchange experiments between calcite and quartz are 1000 in alpha(qz-cc) = 0.38 (+/- 0.06) x 10(6)/T2, far smaller than in the present study. Application of the experimental mineral-water, and especially the direct-exchange calibrations to natural samples, yields unreasonably low geological temperatures. It is difficult to envision a mechanism whereby the measured fractionations in greenschist-grade marbles can be reconciled with the very low temperature estimates obtained with the direct-exchange experimental calibration. Oxygen diffusion rates in quartz are too slow to explain the discrepancy. Postmetamorphic exchange could have occurred with a hydrothermal fluid, but it is unlikely that the delta O-18(calcite) values of all samples would be shifted by an amount that would result in a linear relationship between 1000 in alpha(qz-cc) and T-2. More likely, the discrepancy is due to a kinetic effect in the experiments. The very small fractionations observed in the direct-exchange experiments may have been caused by diffusion-related effects during recrystallization of the quartz and calcite. The problem of recrystallization is eliminated in mineral-CO2 exchange experiments. Combined CO2-calcite and CO2-quartz glass experiments yield the expression 1000 in alpha(qz-cc) = 0.78 (+/- 0.08), in good agreement with the empirical calibration. The new empirical calibration yields reasonable temperature estimates for a wide range of samples and can be used for thermometry in rock types and over temperature intervals where other quantitative geothermometers are lacking.