In this paper, we study the effects of model complexity on the accuracy of the results in the computer simulation of transcranial magnetic stimulation (TMS). The method has been extensively used in the last decade as a noninvasive technique to excite neurons in the brain by inducing weak electric currents in the tissue and proved to be a very promising alternative for currently invasive treatments in Parkinson's and Alzheimer's diseases. A detailed 3-D model of a human head has been developed by combining individual patient-based brain images and the public domain Visible Human data consisting of brain white/gray matter, CSF, skull, and muscles. The finite-element method (low-frequency Ansoft Maxwell 3D package) is used to simulate the interaction of time-varying magnetic fields with brain tissues and to compute the densities of induced currents in different areas. Models with different levels of tissue separation have been developed and tested under the same condition to investigate the effects of model complexity on the distribution of fields and induced currents inside different tissues.