In this article, a statistical-thermodynamic formula based on a new approach has been developed to predict electrical conductivity of carbon-filled composites used for bipolar plate of proton exchange membrane fuel cell. In this model, based on percolation threshold phenomenon, it is assumed that the relationship between electrical conductivity of composite and filler volume fraction follows a sigmoidal equation. Afterwards, the four effective factors on composite conductivity including filler electrical conductivity, filler aspect ratio, wettability, as well as interface contact resistance are replaced upon constant parameters of sigmoidal function. In order to test the model, some single-filler composites have been manufactured by using the phenolic resin as binder and graphite (G), expanded graphite (EG), and carbon fiber (CF) as fillers. The fitting quality is measured by R-square, adjusted R-square, SSE, and RMSE parameters. The results showed that there is a noteworthy agreement between the model and the experimental data. Compared to the other models, this model can be used for more types of fillers. (c) 2012 Wiley Periodicals, Inc. J. Appl. Polym. Sci., 2013