A previously published computer simulation was tested in a biological preparation by recording action potentials from frog sciatic nerves within a volume conductor filled with Ringer's solution. Traveling in a straight line, nerve action potentials traversed a constricted cylinder before crossing into a larger, hemicylindrical volume. Recordings from widely spaced electrodes in the larger volume demonstrated a potential associated with the action potential crossing the boundary between the two volumes. Another potential was associated with the action potential reaching the nerve's cut end. These potentials did not diminish in amplitude with increasing distance from the source. In other recordings, a potential associated with a bend in the nerve was found which was dependent upon the angle of the bend. These results indicate that the simple model of a dipole in a bounded sphere in which potentials decrease as a function of distance from the generator does not explain all potentials that can be observed under conditions that approximate human and animal recordings.