We have compared various induced polarization (IP) models for permeability prediction of the same general form that were all based on two parameters, the first being an electric substitute of effective porosity (the formation factor) and the second being an electric proxy of pore-normalized surface area (the imaginary part of electric conductivity). These models (empirically derived and based on mechanistic formulations) were applied to an extensive database acquired on sandstones and unconsolidated sandy materials. Whereas previous studies on permeability prediction mainly concentrated on either sandstone or unconsolidated sediments, we investigated a database composed of 94 samples including sandstones and unconsolidated material. Most of the samples in the database were saturated with a NaCl solution with an electric conductivity close to . Samples with a saturating fluid that deviated from this composition were corrected using recently published relationships describing the IP dependence on the pore fluid composition. In the case of the sandstone samples, the electric formation factor exerts the primary control on permeability, and the imaginary conductivity was found to be of little importance in permeability prediction. The opposite was observed for the unconsolidated samples, in which the imaginary conductivity was the most important term for permeability and the formation factor was found to be of little importance. The findings suggest that only one property (formation factor in the case of sandstone, imaginary conductivity in the case of unconsolidated samples) might be needed for the order of magnitude estimates of permeability from the popular form of the model based on the IP observations examined here. Whereas the formation factor was challenging to reliably estimate in situ, the imaginary conductivity was directly obtainable from an IP measurement. This suggests that field scale, single-frequency, IP-based estimation of permeability would be challenging, and possibly ineffective, in sandstones.