While many theories of general anesthesia postulate a lipid site of action, there has been no adequate explanation for the lack of anesthetic potency of the highly hydrophobic primary alkanols with more than 12 carbons (the cut-off). Some work suggests that these nonanesthetic alcohols do not dissolve in membranes. Other work contradicts this and suggests that an anesthetic site on a protein provides a better explanation. Here we show that both the anesthetic dodecanol and the nonanesthetic tetradecanol are taken up equally well into the tissues of animals and into isolated postsynaptic membranes. When a group of Rana pipiens tadpoles were treated with dodecanol, half were anesthetized by 4.7 microM (free aqueos concentration), and the corresponding concentration in the tissues was found to be 0.4 mmol per kg wet weight. Prolonged exposure (92 hr) to tetradecanol produced even higher tissue concentrations (0.7 mmol per kg wet weight), yet no anesthetic effects were observed. Furthermore, general anesthetics are thought to act on postsynaptic membranes but both alkanols partitioned into postsynaptic membranes from Torpedo electroplaques. The spin label, 12-doxyl stearate, was incorporated into these membranes. The lipid order parameter it reported was decreased by the anesthetic alcohols (octanol, decanol, and dodecanol), whereas the nonanesthetic alcohols either did not change it significantly (tetradecanol) or actually increased it (hexadecanol and octadecanol). Thus, although lipid solubility is unable to account for the pharmacology of the cut-off in potency of the long-chain alcohols, lipid perturbations provide an accurate description.