Patch-clamp experiments were done on sodium channels of neuroblastoma cells (N1E-115) in the presence of tetraethylammonium ions to block potassium channels. In Ringer solution whole-cell records revealed a diphasic INa inactivation with the fast (τ0) component being clearly larger than the slow (τ1 ≈ 3 τ0) component. In single-channel studies on inside-out patches the mean open time, ¯t0, turned out to be only a fraction of τ0 and almost independent of membrane potential. After external application of chloramine-T INa inactivation of whole cells was delayed with both τ0 and τ1 increased, and incomplete, i.e. a persistent current component emerged. The latter was maximal at a more positive membrane potential than the peak current. Also, after chloramine-T treatment the peak INa increased, particularly at weak depolarizations. In inside-out patches the equally effective internal application of chloramine-T led to bursting channel openings with mean burst times (¯tb) ≈ 6 ms, and gap times (¯tg) ≈ 20 ms, where gap is defined as a closure of ⩾ 1.5 ms. Within the bursts (¯to) was approximately 2 ms, again clearly shorter than τ0; the mean close time, ¯tc was approximately 0.5 ms. The single-channel conductance was approximately 13 pS and unaffected by chlopramine-T. Diphasic INa inactivation and the fact that ¯to < τ0 led to an extension of the model of Aldrich and Stevens [J Neurosci 7:418–431 (1987)], in which overall kinetics is determined by the openings rather than closures of the sodium channels. The extension comprises two inactivated states in series with the open state. Estimates are given of all rate constants of the transition between states that describe the single-channel results as well as whole-cell INa (t), both in the control and in chloramine-T.