Two 3-D electrical models of the human thorax, each consisting of 216000 control volumes, were constructed based upon MR images taken at end diastole and end systole. Using the finite difference method, the contributions of various sources to the impedance cardiogram were studied for the traditional band electrode configuration. The contributions were categorized into three areas: 1) the structural changes between end diastole and end systole, 2) the flow-induced blood resistivity changes in major arteries and veins, and 3) the lung resistivity variation due to the lung blood volume change. Based on the models, Zo and ΔZ between end diastole and end systole were 24.4 Ω and −0.132 Ω, as compared with the measurements of 21.8 Ω and −0.123 Ω made on the same subject from whom the images were taken. Arterial and venous blood resistivity changes caused approximately 57% of the total impedance change. The lung resistivity change and the structural changes contributed 39% and 4%, respectively. The structural changes inside the thorax included the dimensional changes of blood vessels, the blood volume changes of the heart chambers, and heart movement. Although the net impedance change due to the structural changes was relatively small, the individual variation of various factors was large, with significant cancellation occurring. The results suggest that the thoracic impedance cardiographic signal is a mixed representation of many imseparable factors and may not be reliable for the stroke volume calculation. Also, the O-wave, which is clinically observed in various cardiac conditions, may be linked to the diastolic blood flow in the central veins.