We studied artificial ore models that contained galena, pyrite, magnetite, graphite and cryptomelane with the time domain induced polarization technique. The models were mixtures of sand and metallic-type, electronically conductive mineral particles. We varied the volumetric content of the particles, their mineral composition and average grain size, as well as the pore water salinity. Based on the Debye decomposition approach, we obtained relaxation time distributions, which contained peaks. From these distributions, we obtained the total chargeability and the peak relaxation time. We correlated these parameters with the particle mineral composition, grain size, particle content and the pore solution resistivity. We also compared the experimental data with the Wong model prediction, which was unable to explain the entire data set. The above-mentioned correlations, in conjunction with some previously published data, allowed us to formulate a new, semi-empirical model that links (1) the total chargeability with the volumetric content of the particles and the total chargeability of the host matrix and (2) the time constant with the particle mineralogy, the particle radius and the pore solution resistivity.