Backgrounds: Atorvastatin which has a highly variable pharmacokinetic profileoften shows double peaks. However, in most cases it has been regarded as a noise or assay error. In this study we compared 4 different absorption models (zero order, $1^{st}$ order, combined and parallel absorption model) to determine which model best describes the absorption process. Methods: Plasma atorvastatin data of 57 healthy subjects from two bioequivalence studies were used. Concentrations were measured up to 48 hours after drug administration. Mixed effect analysis to compare the 4 absorption models was performed using NONMEM version 6.0. Results: The absorption process was best described bycombined absorption model. CL/F and Vd/F were 400L/hand 779L, respectively, and KA and D were $0.09;hr^{-1}$ and $0.53;hr^{-1}$, respectively. Conclusion: Although the combined absorption model did not fully describe the double peaks, this study demonstrated that absorption models other than simple first or zero order absorption are necessary to describe the double peaks phenomenon.