Conduction in demyelinated and remyelinating axons has been simulated with a computational model. The calculations made use of recent determinations of ionic channel densities in the internodal axolemma of Xenopus fibers. Several new morphological measurements reduced the number of parameters not directly obtained from experimental data. Action potentials and ionic currents were calculated for a wide range of fiber diameters and internodal lengths. The earliest stage of remyelination, characterized by Schwann cell attachment and extension of processes, was simulated by covering just a small percentage of the internode by a single cell layer. Conduction invariably failed if the internodal Na+ channel density was zero. The minimum density required for successful propagation agreed well with that measured in loose patch clamp experiments. Lateral diffusion of Na+ channels from nodes of Ranvier into the demyelinated internode did not restore conduction in blocked axons, and this was true regardless of the initial internodal Na+ channel density. Decreases in the internodal K+ channel density improved the safety factor for conduction, but this was significant only in the largest axons. Simulating minimal paranodal demyelination by eliminating the axo-glial junctional seals did not result in conduction block, but did produce large conduction delays.