SUMMARYA method is described for the measurement of narcotic potencies of chemical compounds, using houseflies. Narcotic potency measurements give rise to statistical problems which differ from those encountered in toxicity determinations in so far as the narcotic process is reversible, and successive values of the number of affected insects in a group subjected to the test do not normally increase monotonically and in some case considerable oscillation occurs. An earlier method (Bradbury&Armstrong, 1954) used probit analysis, but it is now shown that this treatment and also regression analysis both fail to provide a valid method of dealing with variability in this type of test. A test based on the response of two flies, one male and one female per flask, is now described. The test, based on sixteen flies per compound, gave figures for:α0, the partial saturation of the test compound giving the same average knock‐down time as chloroform at a partial saturation of 0.005; α1, the partial saturation of the test compound giving knock‐down in I hr.;b, the slope of the straight line relating log. knock‐down time to log. partial saturation;s, the residual standard deviation of log. knock‐down time about its regression line on log. partial saturation.From these values the 95% confidence limits for the slope and for the values of α0were calculated. The test also provided information on the convulsive action of compounds and a modification of it was used to determine knock‐down times of very potent narcotics at high dosages. Results are presented for fifty‐seven halogenated hydrocarbons and the hypothesis that variations in potency between narcotics are determined mainly by rates of diffusion is examined. The results do not support this hypothesis. Narcotic and convulsant potencies measured by the method described are a useful guide to the behaviour of compounds