The validity of a one- and a two-cylinder model, underlying thoracic impedance cardiography (TIC), was investigated by studying the length dependence of the impedance parameters Z0, (dZ/dt)min, and stroke volume (SV). It can be shown that, within a one-cylinder model, all parameters are directly proportional to the length, whereas, if the volume conduction of the thorax and the neck are modeled separately, Z0 and (dZ/dt)min are expected to be linear dependent and SV will be nonlinear upon the length. The expectations were compared to results from in vivo measurements. Two electrode arrays were studied, in which the caudal recording electrode position was varied; SV was calculated using Kubicek's equation. Except for small distances, the results showed a nearly linear relation between the parameters and the length. Regression analysis of the linear part revealed statistically significant intercepts (p < 0.05). Neither the intercept nor the nonlinear part can be explained by a one-cylinder model, whereas a model consisting of two cylinders serially connected describes the experimental results accurately. Thus SV estimation based on a one-cylinder model is biased due to the invalid one-cylinder model. Corrections for the Kubicek-equation need to be developed in future research using this two-cylinder model.