The objective of this work is to derive the soil water retention from the soil structure without curve-fitting and only using the physical parameters found irrespective of an experimental retention curve. Such physical modeling is more preferable for both data analysis and prediction. Two key points underlie the work: (i) the soil suction at drying coincides with that of the soil intra-aggregate matrix and contributive clay; and (ii) both the soil suction and volume shrinkage at drying depend on the same soil water content. In addition the two following results are used: (i) the available two-factor (capillarity and shrinkage) model of clay suction enables one to connect a clay suction and clay water content using the clay matrix structure; and (ii) the recent reference shrinkage curve model based on the concepts of intra-aggregate soil structure permits one to connect the soil water content at shrinkage with the water content of the contributive clay. With that the available two-factor model was essentially modified and, in particular, the effect of adsorbed water film was taken into account. The developed model includes the following input parameters: the solid density, relative volume of contributive- clay solids, relative volume of contributive clay in the oven-dried state, soil clay content, aggregate/intra-aggregate mass ratio, and specific volume of lacunar pores in the aggregates at maximum swelling. The validation of the model is based on available data of water retention and the above input parameters for six soils. A promising agreement between the predicted and observed water retention curves was found.