Human head models for the forward computation of EEG using FEM require a large set of elements to represent the head geometry accurately. Anatomically, the electrical property of each element is different, even though they may represent the same type of tissue (white matter, grey matter, etc.). Since it is impossible to obtain the electrical properties of the cranial tissues for every element in the head model, most algorithms which claim to deal with inhomogeneity can, in reality, only implement the computation for the homogeneous case. This paper presents a new numerical approach which can more precisely model the head by using a set of pseudo conductivity values for the computation of the inhomogeneous case. This set of values are extrapolated from the limited amount of real conductivity values which are available in the literature. Simulation studies, based on both this proposed approach and the homogeneous approach which utilizes mean-valued conductivities, are performed. The studies reveal that the computation results for the potential distribution on the surface of the scalp, obtained using both approaches, are significantly different.