Saturated hydraulic conductivity, Ks, is an essential parameter for predicting transport processes through soils. Although relatively easy to measure, Ks is a property that is not routinely measured during soil surveys. Therefore, it is desirable to have predictive models based on available soil properties (texture, bulk density, particle size distribution and porosity) to facilitate modeling efforts at a regional or global scale. Pedotransfer functions of Ks are notoriously inaccurate. An alternative approach is to develop a physically sound model containing easy to predict parameters. The task of developing such model has been simplified by the recent compilation in the USA and Europe of large databases containing basic soil properties and Ks values. Among the most successful Ks models are those that use a characteristic particle diameter or effective porosity raised to a constant or variable power and modified by a linear constant. One such model is the one proposed by Rawls et al. (Trans. ASAE 41: 983-988; 1998) that uses effective porosity (defined as air-filled porosity at a matric potential of -33 kPa) modified by a power representing a pore distribution index (defined as the change in porosity between -33 kPa and -1,500 kPa) and by an empirical linear constant . Our hypothesis is that the prediction of Ks can be improved by using information of water retention at relatively low matric potentials (indicative of macroporosity). A recently proposed theory that relates soil physical quality to the inflection point of water retention curves can be used to generate information at low matric potentials, which is generally absent in large databases. Consequently, our objective was to modify Rawls et al. (1998) model by redefining effective porosity, the power (pore distribution) index, and the linear constant with a combination of parameters derived from water retention information at low matric potentials. Twenty-three sandy materials from New Jersey were sampled and measured for particle-size distribution, water retention curves, and Ks. The same properties of 213 New Jersey horizons, 77 horizons from UNSODA database, 13 rootzone mixes, and 48 Alabama horizons were added to the original database. Water retention curves were fitted with the van Genuchten (Soil Sci. Soc. Am. 44: 892-898, 1980) model and the parameters of the model used to analytically determine the water retention inflection points. Modifications introduced to the Rawls et al. (1998) model included: 1) defining a new effective porosity (as air-filled porosity at the inflection point) and power index (change is porosity between saturation and the inflection point), and 2) expressing the linear constant as a function of the pore diameter at inflection point and of the power index. All samples were grouped into 17 textural classes. The coefficient of correlations between the measured values of Ks and those predicted with the proposed model was R² = 0.95 when comparing the 17 textural averages, and R² = 0.72 when comparing all 374 values. Further testing of the model with 1,729 data points (grouped in 11 textural classes) from the European HYPRES database produced similar results (R² = 0.98). The residuals of the fitting were randomly distributed indicating no apparent bias in the prediction. The proposed model is physically sound, contains no fitting parameters, and uses information that can be obtained from pedotransfer functions. With site specific data the model has the potential to be used in investigations at the field scale.