Accurate analyses of large numbers of soil samples are needed in order to reduce the uncertainty of carbon inventories. Loss-on-ignition (LOI) is still considered the most convenient assessment method, but its accuracy and precision for predicting total organic carbon (TOC) is questioned. However, our estimation of measurement precision for different samples showed comparable relative standard deviations (RSDs) for LOI and TOC determinations. Highest precision was found in forest floor samples (RSD <1.2%) and lowest (RSD 5-10%) in sandy soil samples low in organic matter. Forest floor samples (n = 66) and non-calcareous mineral soil samples (n = 654) were used to calibrate and validate predictive equations. Excellent linear relationships were found. For a wide range of soils the bivariate predictive equation TOC = -0.1046 Clay + 0.5936 LOI (r2 = 0.98) was developed and validated. After correction for clay content, slopes averaged remarkably close to the traditional 0.58 conversion factor. Copyright © Taylor & Francis, Inc.