We investigated the precision of spike timing in a model of gap junction-coupled oscillatory neurons. The model incorporated the known physiology, morphology, and connectivity of the weakly electric fish's high-frequency and extremely precise pacemaker nucleus (Pn). Two neuron classes, pacemaker and relay cells, were each modeled with two compartments containing Hodgkin-Huxley sodium and potassium currents. Isolated pacemaker cells fired periodically, due to a constant current injection; relay cells were silent but slightly depolarized at rest. When coupled by gap junctions to other neurons, a model neuron, like its biological correlate, spiked at frequencies and amplitudes that were largely independent of current injections. The phase distribution in the network was labile to intracellular current injections and to gap junction conductance changes. The model predicts a biologically plausible gap junction conductance of 4-5 nS (200-250 MOmega). This results in a coupling coefficient of approximately 0.02, as observed in vitro. Network parameters were varied to test which could improve the temporal precision of oscillations. Increased gap junction conductances and larger numbers of cells (holding total junctional conductance per cell constant) both substantially reduced the coefficient of variation (CV = standard deviation/mean) of relay cell spike times by 74-85% and more, and did so with lower gap junction conductance when cells were contacted axonically compared with somatically. Pacemaker cell CV was only reduced when the probability of contact was increased, and then only moderately: a fivefold increase in the probability of contact reduced CV by 35%. We conclude that gap junctions facilitate synchronization, can reduce CV, are most effective between axons, and that pacemaker cells must have low intrinsic CV to account for the low CV of cells in the biological network.