Soil water retention is an important hydraulic property in the study of water flow and solute transport in soils. However, soil water retention measurements are costly and time-consuming. In this study, a procedure was developed to estimate this function based on soil particle-size distribution and the fractal theory. A relationship between the fractal dimension and cumulative particle-size distribution was derived. Based on the relationship, the fractal dimension was determined using particle-size distribution data. The fractal dimension was incorporated as the exponent parameter of retention models. The procedure was tested using particle-size distribution data and soil water retention data of 110 soils from a data base. Good agreement between the experimental data and estimations of soil water retention from the procedure was obtained for most of the soils. The results suggest that the procedure is a sufficiently accurate and economic method to predict the soil water retention function.