The study of compound action potentials has not succeeded in determining exact limits to differential block of nerve fibers by local anesthetics. Further observations on individual neurons therefore were undertaken. Rabbit vagus nerve and ganglion were superfused in vitro at 37 degrees C, pH 7.4. Activity evoked by stimulating the distal end of the nerve was monitored extracellularly from individual somata by a microelectrode in the ganglion. Lidocaine HCl in steps of 0.2-0.1 mmol/l (0.0005-0.0025 g/dl) was applied to the intervening nerve to identify two concentrations, respectively, just sparing and extinguishing conduction in the axon belonging to the soma, the average being regarded as the blocking concentration. The mean blocking concentrations (mM, +/- SD) for axons conducting at control velocities between 3 and 26 m/s (myelinated axons) was 0.43 +/- 0.15, N = 18, for axons slower than 1.4 m/s (unmyelinated axons) 0.63 +/- 0.14, N = 11, for intermediate velocity axons 0.19 +/- 0.18, N = 7. The differences were significant by Scheffé's multiple comparisons test (P less than 0.01). Although the myelinated axons were blocked by a lower average concentration of lidocaine than unmyelinated axons, they manifested significantly more slowing of conduction before block (P less than 0.001). No relation between blocking concentration or latency increase and conduction velocity (fiber size) was evident within any fiber group.