The potential influence of the transverse heterogeneity in the sinusoidal enzyme contents and the capillary transit times on the elimination from the liver sinusoids was evaluated by using distributed models. Moreover, the predictions of the undistributed model (the parallel tube model), the venous equilibration model (the well-stirred model), the new distributed models presented in this study and the distributed model by Bass et al. (J. Theor. Biol., 72, 161 (1978)) for the hepatic drug clearance were compared. Data on lidocaine kinetics reported by Pang and Rowland (J. Pharmacokinet. Biopharm., 5, 655 (1977)) were employed. The steady state output concentration of lidocaine in the blood leaving the liver was predicted better by the well-stirred and distributed models than by the parallel tube model. Because both the parallel tube and well-stirred models ignore the anatomic and metabolic heterogeneities in the liver lobule, they may be refuted experimentally and theoretically, as reported by Goresky et al. (J. Clin. Invest., 52, 991 (1973)). It is considered that the distributed model taking account of the transverse heterogeneity in the sinusoidal enzyme contents and the capillary transit time, based on anatomical evidence, can well accommodate the experimental data on lidocaine.