Inactivation of sodium channels was studied through gating current (Ig) experiments. Ig, a small and slow component of capacitative current, is caused by molecular rearrangements in ionic channels. While no component of Ig has a time course similar to inactivation, inactivation affects Ig by immobilizing about two-thirds of gating charge. Immobilization can be measured by comparing ON and OFF charge movement during a pulse. The OFF:ON ratio is near 1 for short pulses and drops to about one-third for longer pulses, following the time course of inactivation. Inactivation and immobilization share the same voltage dependence and recovery kinetics. At -150 mV, immobilized charge forms a slow component of current as it returns to the off position. Pronase treatment, which destroys inactivation, removes immobilization. A model is proposed in which inactivation gains its voltage dependence by coupling to the activation gate. Inactivation immobilizes two-thirds of gating charge, which returns slowly to the off position. At -150 mV, this return is rapid enough to produce a slow OFF component. The slow OFF component is reduced by pronase, which destroys inactivation. The slow OFF component develops with the time course of inactivation, as shown by fitting tails with two exponentials. At -140 or -150 mV, recovery from inactivation is fast enough to allow immobilized charge to return to the off position, contributing to the slow OFF component. The slow OFF component is associated with the return of immobilized charge and is not present in pulses from -70 mV. Inactivation alters the Q-Vm distribution by immobilizing charge. The Q-Vm distribution remains unchanged during activation but changes during inactivation. Inactivation of Ig can be demonstrated by other pulse patterns, showing that inactivation affects both Ig and INa. Na channels may have three open states: x1, x1y, and h2. The slow ON component of Ig is not correlated with any known process, but is related to the transition from x1 to x1y. The model accounts for various experimental findings, including delayed onset of inactivation, slow inactivation for small depolarizations, voltage-dependent inactivation, pronase effects, charge immobilization, and variation of closing kinetics with pulse duration. The model also satisfies the condition of microscopic reversibility, requiring that forward and backward fluxes be equal. The model is supported by pharmacological studies showing that inactivation of Na and K channels is similar. The model predicts current fluctuations, or noise, distinct from the Hodgkin and Huxley predictions.Inactivation of sodium channels was studied through gating current (Ig) experiments. Ig, a small and slow component of capacitative current, is caused by molecular rearrangements in ionic channels. While no component of Ig has a time course similar to inactivation, inactivation affects Ig by immobilizing about two-thirds of gating charge. Immobilization can be measured by comparing ON and OFF charge movement during a pulse. The OFF:ON ratio is near 1 for short pulses and drops to about one-third for longer pulses, following the time course of inactivation. Inactivation and immobilization share the same voltage dependence and recovery kinetics. At -150 mV, immobilized charge forms a slow component of current as it returns to the off position. Pronase treatment, which destroys inactivation, removes immobilization. A model is proposed in which inactivation gains its voltage dependence by coupling to the activation gate. Inactivation immobilizes two-thirds of gating charge, which returns slowly to the off position. At -150 mV, this return is rapid enough to produce a slow OFF component. The slow OFF component is reduced by pronase, which destroys inactivation. The slow OFF component develops with the time course of inactivation, as shown by fitting tails with two exponentials. At -140 or -150 mV, recovery from inactivation is fast enough to allow immobilized charge to return to the off position, contributing to the slow OFF component. The slow OFF component is associated with the return of immobilized charge and is not present in pulses from -70 mV. Inactivation alters the Q-Vm distribution by immobilizing charge. The Q-Vm distribution remains unchanged during activation but changes during inactivation. Inactivation of Ig can be demonstrated by other pulse patterns, showing that inactivation affects both Ig and INa. Na channels may have three open states: x1, x1y, and h2. The slow ON component of Ig is not correlated with any known process, but is related to the transition from x1 to x1y. The model accounts for various experimental findings, including delayed onset of inactivation, slow inactivation for small depolarizations, voltage-dependent inactivation, pronase effects, charge immobilization, and variation of closing kinetics with pulse duration. The model also satisfies the condition of microscopic reversibility, requiring that forward and backward fluxes be equal. The model is supported by pharmacological studies showing that inactivation of Na and K channels is similar. The model predicts current fluctuations, or noise, distinct from the Hodgkin and Huxley predictions.