To measure the depth of the local anesthetic binding site within the neuronal membrane, biotin-containing polyethylene glycols having zero, three, and six ethylene glycol subunits were added to the p-amino termini of tetracaine and procaine, thereby interposing a pharmacologically inert "spacer" molecule between the local anesthetic and the biotin moiety. These biotinyl-local anesthetic derivatives produced "tonic" inhibition of the compound action potential of split, desheathed frog sciatic nerves in a concentration-dependent, reversible manner. However, no inhibition of the action potential occurred when sufficient avidin, a 66,000-MW protein that binds four biotins, was present to bind and anchor the biotin-containing end of each derivative outside the plasma membrane. Increasing the "leashed" anesthetic derivative's concentration to 4 times that which reduced impulse height by 50% in the absence of avidin still produced no detectable block when equimolar avidin was present. Apparently, the "spacer" in the derivative compound was too short to permit the avidin-complexed anesthetic to reach its site of action on the sodium channel. In a similar fashion, the local anesthetic derivatives produced "use-dependent" block when drug-treated nerves were stimulated at 40 Hz in the absence of equimolar avidin, but failed to produce "use-dependent" block when equimolar avidin was present. In common with others, we assume that tertiary amine local anesthetics may reach their binding site via hydrophobic (transmembrane) pathways without necessarily entering the cytoplasm. Thus, since our longest local anesthetic derivative, that containing six ethylene glycol subunits, placed the local anesthetic group a maximum of 15-18 A from the surface of the avidin moiety, we conclude that the local anesthetic binding site for block of sodium channels of amphibian nerve must be greater than or equal to 15 A from the outer surface of the plasma membrane.