The localization of voltage-sensitive sodium channels was investigated in mouse, rat and rabbit sciatic nerves using iodinated alpha- and beta-Scorpion toxins (ScTx) as specific probes. Saturable specific binding for a beta-ScTx was detected in mouse sciatic nerve homogenates (Kd = 90 pM, binding site capacity = 90 fmol mg-1 protein). LM autoradiographic studies demonstrated that the two types of ScTx stained the Ranvier nodes of the myelinated fibres, and also showed a clear but weaker labelling of the unmyelinated Remak bundles. In the sciatic nerve, which is widely considered as a model 'myelinated nerve', the nodal membrane represented only a small fraction of the total axonal membranes (0.2% and 0.05% for mouse and rabbit sciatic nerves respectively). Therefore, despite their high channel density, nodal membranes contribute only a small proportion of the total labelling by beta-ScTx (15% and 2.3% for mouse and rabbit sciatic nerves respectively), with the major contribution to labelling arising from unmyelinated axons. The distribution of specific binding sites for a beta-Scorpion toxin was then analysed in cross-sections of rabbit sciatic nerve at the EM level. The quantitative analysis of autoradiograms involved three methods, the 50% probability circle method, and two cross-fire analyses using either systematically distributed hypothetical sources or hypothetical sources only located on the plasma membranes of axons and of Schwann cells associated with unmyelinated Remak bundles. No specific beta-Scorpion toxin binding sites were detected at the plasma membrane of Schwann cells from either myelinated fibres or unmyelinated bundles, or at the internodal surface of myelinated axons. Sites were only detected at the surface of unmyelinated axons and at nodal axolemma. Their density in unmyelinated axons was found to be in the range of 1-6 per micron2 of plasma membrane surface area by combining quantitative EM autoradiography and stereological measurements.