In a previous report, the incidence of tourniquet pain was found to be 25% with bupivacaine and 60% with tetracaine (P less than 0.05) spinal anesthesia. On the other hand, tetracaine is more potent than bupivacaine in abolishing the single-compound action potential in vitro in isolated nerves. These conflicting observations may be reconciled if bupivacaine produced greater frequency-dependent conduction blockade of nerve action potentials. This hypothesis was tested in C fibers of isolated, desheathed rabbit vagus nerves. The nerves were supramaximally stimulated at frequencies of 9 or 15 Hz. After a control period, the nerves were exposed to bupivacaine (0.2 mM) or tetracaine (0.02 mM) for 30 minutes. The local anesthetics were then washed out by continuous constant-rate perfusion. The decline and recovery of the first and last action potential amplitudes of the train were measured. Bupivacaine and tetracaine produced similar depression of the first action potential of the 9-Hz and 15-Hz trains. However, bupivacaine caused a delayed recovery of the last action potential of the 15-Hz train but not of the 9-Hz train. These results show that bupivacaine produces greater frequency-dependent conduction blockade of C fibers than does tetracaine. These findings offer a possible explanation as to why spinal anesthesia with bupivacaine results in a lower incidence of tourniquet pain than tetracaine.