We used two conditioning voltage protocols to assess inactivation of voltage-gated Na+ current in retinal ganglion cells. The first protocol tested the possibility, raised by published activation and steady-state inactivation curves, that Na+ ions carry a "window" current in these cells. The second protocol was used, because these cells spike repetitively in situ, to measure the Na+ current available for activation following spikes. Na+ current activated at test potentials more positive than –65 mV. At test potentials more positive than –55 mV, Na+ current peaked and then declined along a time course that could be fit by the sum of a large, rapidly decaying component, a small, slowly decaying component and a non-decaying component. Both step- and spike-shaped conditioning depolarizations reduced the amount of current available for subsequent activation, sparing the non-decaying "persistent" component. Most of the Na+ current recovered from this inactivation along a rapid exponential time course (τ=3 ms). The remaining recovery was complete within at least 4 s (at –70 mV). Our use of step depolarizations has identified a current component not anticipated from previous measurements of steady-state inactivation in retinal ganglion cells. Our use of spike-shaped depolarizations shows that Na+ current density at 1 ms after a single spike is roughly 25% of that activated by the conditioning spike, and that recovery from inactivation is 50–90% complete within 10 ms thereafter. Na+ current amplitude declines during spikes repeated at relatively low frequencies, consistent with a slow component of full recovery from inactivation.