Several models have been developed to estimate the water retention curve (WRC) on the basis of the similarity in shape between the WRC and the particle-size distribution (PSD) of a soil. Shortcomings of these models are (i) their adoption of empirical equations for the PSD and WRC and (ii) use of regression equations for the WRC parameters that do not emphasize the physical significance of their model parameters. By applying a lognormal distribution law to the PSD and assuming a lognormal void-size distribution (VSD) and then combining with the WRC function developed by Kosugi in 1996, we developed a simple model for estimating the WRC directly from the PSD. The only parameters that need to be estimated are those defining relationships between particle diameter (d) and pore size (r). Data for 117 sandy soils were extracted from the UNSODA database. The parameters were calibrated for three texture classes within sandy soils used in this study. The model defining the nonlineard-rrelationship (NL model) predicted the WRC better than the model defining the lineard-rrelationship (L model). The L model overpredicted the capillary pressure at high saturations, whereas the opposite was observed at the low saturation region. The inclusion of two parameters that describe the packing characteristic as well as the respective shapes of the particles and the corresponding pores may explain the better performance of the NL model.