Electrical resistance tomography (ERT) can be used for the noninvasive characterization of soil moisture and soil structural heterogeneity. Any attempt to relate electrical resistivity measurements to soil moisture content or soil bulk density, however, must rely on a “pedo-electrical” function, i.e., a conductivity model for soils. This study aimed to test five pedo-electrical models for their ability to reproduce electrical resistivity as measured by ERT in a silt loam soil sample across a range of moisture and bulk density values. The Waxman and Smits model, the Revil model, the volume-averaging (VA) model, the Rhoades model, and the Mojid model were inverted within a Bayesian framework, thereby identifying not only the optimal parameter set but also parameter uncertainty and its effect on model prediction. The VA model outperformed the other models in terms of both fit and parameter consistency with respect to independent estimates of surface conductivity obtained with published pedotransfer functions. Sensitivity of the electrical resistivity was then studied by means of the calibrated VA model, revealing an approximately 1.5 times higher sensitivity to soil moisture content than to soil bulk density. In addition, the sensitivity of electrical resistivity to soil moisture and soil bulk density was found to increase as soil moisture and bulk density decreased. The VA model calibrated on the basis of resistivity measurements appeared to simulate relatively well the measured soil moisture content for electrical resistivity values <100 Ω m. As opposed to water content, the soil porosity was badly approximated by the model. It appears therefore that ERT is more suitable for detecting heterogeneity in soil water content than differences in soil bulk density.