Phasic block of rat cardiac Na+ current by saxitoxin was assessed using pulse trains and two-pulse voltage clamp protocols, and the results were fit to several kinetic models. For brief depolarizations (5 to 50 ms) the depolarization duration did not affect the rate of development or the amplitude of phasic block for pulse trains. The pulse train data were well described by a recurrence relation based upon the guarded receptor model, and it provided rate constants that accurately predicted first-pulse block as well as recovery time constants in response to two-pulse protocols. However, the amplitudes and rates of phasic block development at rapid rates (> 5 Hz) were less than the model predicted. For two pulse protocols with a short (10 ms) conditioning step to -30 mV, block developed only after repolarization to -150 mV and then recovered as the interpulse interval was increased. This suggested that phasic block under these conditions was caused by binding with increased affinity to a state that exists transiently after repolarization to -150 mV. This "post-repolarization block" was fit to a three-state model consisting of a transient state with high affinity for the toxin, the toxin bound state, and the ultimate resting state of the channel. This model accounted for the biphasic post-repolarization block development and recovery observed in two-pulse protocols, and it more accurately described phasic block in pulse trains.(ABSTRACT TRUNCATED AT 250 WORDS)