Measurements of stable carbon isotopic composition (d13C) of carbonates or carbonate-rich soils are seldom performed in a continuous-flow isotope ratio mass spectrometer (IRMS) using an elemental analyzer (EA) as an online sample preparation device. Such analyses are routinely carried out with an external precision better than 0.1% using a GasBench II (GB) sample preparation device coupled online with a continuous-flow IRMS. In this paper, we report and compare d13C analyses (86 total analyses) of calcium carbonates obtained by using both the GB and the EA. Using both techniques, the d13C compositions of two in-house carbonate standards (MERCK carbonate and NR calcite) and ten selected carbonate-rich paleosol samples (of variable CaCO3 content) were analyzed, and data are reported in the VPDB scale calibrated against international standards, NBS 18 and 19. For the inhouse standards analyzed by both techniques, a precision better than 0.08% is achieved. The analytical errors (1s) computed from multiple analyses of the d13C of both the MERCK and NR obtained by the above two techniques are nearly identical. In general, the 1s (internal error) of paleosol analyses obtained in the GB is better than 0.06%, whereas that for the analyses in the EA (three repetitive analyses of the same sample) varies in the range 0.05–0.21%. However, for paleosols having more than 85% CaCO3, 1s is better than 0.15%(similar to the instrument precision), and in this case the d13CVPDB of samples obtained by the GB is similar to that obtained by the EA. Our results suggest that the d13C of pure calcium carbonate samples can also be analyzed using the EA technique. Copyright ©2006 John Wiley & Sons, Ltd.