We examined the dependence of imaginary conductivity () on pore fluid conductivity () for an extensive database of 67 samples acquired from twelve independent studies. We compared fitting of functions describing the salinity dependence of for two models of the electrical double layer (EDL) polarization, both of which predict asymptotic behavior of at high . We define these models as the diffuse layer polarization (DLP) and Stern layer polarization (SLP) models based on the physical description of the salinity dependence of the surface polarization. We also examined the database for evidence of a high salinity decrease in not predicted by either model. The dependence of on prior to the polarization plateau predicted by both models approximates a simple empirical power law with an average exponent of 0.34. The salinity dependence predicted by the DLP model adequately describes most data sets. A fitting parameter representing the high salinity asymptote is strongly correlated () with pore normalized specific surface (). The SLP model describes well the observations when a recently proposed additive polarization term representing the contribution of the protons is included. In this case, the SLP model provides an excellent fit to the data sets, including a low salinity asymptote (in log-log conductivity space) seen in some samples. Predicted values of the fitting parameters of the SLP model generally are consistent with the values expected based on the theory; the fitting parameter describing the high salinity asymptote of the SLP model is also strongly correlated () with . The SLP and DLP models neglect a high salinity decrease in the polarization that is observed in numerous data sets from independent studies. New data acquired on a sandstone sample demonstrate that this high salinity decrease is likely not attributable to the limited phase accuracy of earlier measurements.