1. We have redetermined the anaesthetic potencies (EC50S) for a series of primary alkanols, to resolve uncertainties about the molecular dimensions of the anaesthetic site resulting from the use of data from different laboratories. 2. For each alkanol, concentration-response relationships for loss of righting reflex (LRR) were plotted for over one hundred tadpoles, and the median effective concentrations determined. Aqueous concentrations present during potency assays were determined independently, and for alkanols with chain length greater than nonanol, correction was made for depletion from the aqueous phase. 3. The EC50S were found to decrease logarithmically with increasing number of carbon atoms in the hydrocarbon chain of the alkanol (CN), such that, on average, each additional methylene group was associated with an approximately four fold increase in potency. 4. The relationship between log EC50 and CN was best described by the quadratic equation, log EC50 = 0.022 (+/- 0.0038) CN2 + 0.76 (+/- 0.051) CN + 3.7 (+/- 0.14) (r2 = 0.9951). 5. A previously described correlation between the apparent changes in the free energy of binding of an additional methylene group both to luciferase and to the sites for LRR in tadpoles was not confirmed. 6. A cut-off in potency beyond dodecanol was established in experiments where tadpoles were maintained in supersaturated solutions of tridecanol for 20 h without demonstrable LRR. 7. These findings indicate that the soluble enzyme firefly luciferase does not adequately model the anaesthetic site. Specifically, there are discrepancies in the position of cut-off, and the apparent changes in the free energy of binding, per methylene group, of an alkanol to luciferase do not parallel that for tadpoles.