Lidocaine block of cardiac sodium channels was studied in voltage-clamped rabbit purkinje fibers at drug concentrations ranging from 1 mM down to effective antiarrhythmic doses (5-20 muM). Dose-response curves indicated that lidocaine blocks the channel by binding one-to-one, with a voltage-dependent K(d). The half-blocking concentration varied from more than 300 muM, at a negative holding potential where inactivation was completely removed, to approximately 10 muM, at a depolarized holding potential where inactivation was nearly complete. Lidocaine block showed prominent use dependence with trains of depolarizing pulses from a negative holding potential. During the interval between pulses, repriming of I (Na) displayed two exponential components, a normally recovering component (tauless than 0.2 s), and a lidocaine-induced, slowly recovering fraction (tau approximately 1-2 s at pH 7.0). Raising the lidocaine concentration magnified the slowly recovering fraction without changing its time course; after a long depolarization, this fraction was one-half at approximately 10 muM lidocaine, just as expected if it corresponded to drug-bound, inactivated channels. At less than or equal to 20 muM lidocaine, the slowly recovering fraction grew exponentially to a steady level as the preceding depolarization was prolonged; the time course was the same for strong or weak depolarizations, that is, with or without significant activation of I(Na). This argues that use dependence at therapeutic levels reflects block of inactivated channels, rather than block of open channels. Overall, these results provide direct evidence for the "modulated-receptor hypothesis" of Hille (1977) and Hondeghem and Katzung (1977). Unlike tetrodotoxin, lidocaine shows similar interactions with Na channels of heart, nerve, and skeletal muscle.