Laboratory and field methods that are used to determine the soil water characteristic curve (h-θ function) are expensive and time consuming. A physical model for predicting the h-θ function based on the soil particle-size distribution curve and soil bulk density with different scale parameters (α, i.e., constant, linear, and logistic models) has been proposed in the literature. Unfortunately, many databases do not contain the full particle-size distribution, but instead contain only the sand, silt, and clay mass fractions. A method for estimating the particle-size distribution from clay, silt, and fine plus very fine sand mass fraction (particle radii, between 25 and 125 μm) has been presented in the literature and is improved by using all sand particles (radii between 25 and 999 μm, modified model). The objectives of the present study were to evaluate the predicted soil water characteristic curve for 16 soil samples with different textures based on clay, silt, and sand fractions, and soil bulk density using the improved method for the prediction of particle-size distribution (modified model), different values for the scale parameter as constant, being obtained from linear and logistic models. The results indicated that for clay and silt loam soils using a radius of 999 μm (the modified model), the particle-size distribution, and consequently soil moisture characteristic curves, was predicted more accurately than those obtained by using a radius of 125 μm for the largest particles of the soil. This is specifically shown for a determined by the logistic procedure. The values of α based on the logistic model were best suited for the clay, silt loam, and loam soils. The values of α based on the linear model were appropriate for the clay and silt loam soils. Further, the values of constant α were best suited for the clay soils. However, these results are considered indicative rather than conclusive due to the small size of the data set for clay, silt loam and sandy loam soils. Finally, it is proposed to test the modified model for a wider data set.