Myelinated nerve fibers exhibit a complex anatomy in the nodal region which includes a marked nodal-paranodal constriction and an intricate paranodal structure where the myelin sheath is separated from the axon by a narrow periaxonal space. In this study, a recently developed computational model of the mammalian myelinated nerve fiber based on electron microscopic data was employed to examine the effect of the nodal-paranodal axonal radius and periaxonal space width on the conduction of action potentials. These findings indicate that the nodal-paranodal constriction promotes higher conduction velocities by minimizing the component of the nodal capacity contributed by the paranodal axolemma. Model prediction of optimal nodal-paranodal radii is correlated with radii determined in experimental anatomical studies.