Soil water retention measurements are relatively time-consuming and become impractical when hydrologic estimates are needed for large areas. One approach to soil water retention estimation from readily available data is based on the hypothesis that soil water retention is additive and can be approximated by summing up water retention of pore subspaces related to the soil components. Our objective was to test this hypothesis. In the 'additivity' model used in this work, the main assumptions are: (a) the additivity is applicable to gravimetric water contents, (b) water retention related to a textural fraction can be measured on packed samples consisting exclusively of this fraction's particles, (c) water retention of textural fractions contributes to total soil water retention in proportion to the volumes of pore subspaces related to each fraction. Water retention of packed cores consisting of soil textural fractions was taken from literature. The resultant model did not have fitting parameters. It was tested with samples of sandy soils that were collected in the UNSODA database from seven states of the United States and from seven other countries. The median root-mean squared error (RMSE) constituted 0.017 m3m−3, 0.023 m3m−3, and 0.028 m3m−3in coarse, medium, and fine sands, respectively. These RMSE are at the lower end of the RMSE range for the regression-based pedotransfer functions found in literature. Slopes and intercepts of the regressions of estimated water contents on measured water did not differ significantly from one and zero at the 0.05 significance level except in one case. The RMSE of the laboratory water retention estimates with additivity model was significantly less than the RMSE of the field water retention estimates from the laboratory water retention data. We discuss possible sources of errors and conclude that the attained accuracy warrants testing the additivity model with other soil data and improving this model to accommodate various types of soil structure and additional soil components.