A major problem confronting application of impedance techniques to studies of disease is the extraction of intrinsic properties of the tissue from the measured impedances, which unavoidably involve geometric factors as well. Amongst the foremost are the sizes and locations of the measuring electrode arrays, and this paper addresses one of these, the location of current injecting electrodes. Tetrapolar impedance measurements on a 17.5 cm segment of the thigh gave R and X values three to four times larger when the current injecting electrodes were placed 2.5 cm from the sensing electrodes than when very distant placement was used. The frequency dependences of R and X were affected as well, though the X versus R plots still showed virtually perfect depressed-center semicircles, as in the Cole model. R(f) and X(f) for the set of contiguous 2.5 cm wide sub-segments show that these behaviors can be explained by a combination of the transverse orientation of current flow lines near the injecting electrodes and the anisotropy of the resistivity associated with the bundled fiber structure of muscle tissue. The measured impedance was found to be a separable function of geometric and intrinsic tissue variables, but far more complicated than is implicit in the usual cylindrical models. The results also suggest that many full and segmental body composition studies in the literature may be prone to substantial errors due to too close placement of the current injecting electrodes.