Chemicals applied to annual weeds frequently gave dosage-mortality curves that were not of the usual sigmoid shape. These departures appear to be attributable to the method of application, as the spray is only partly retained by the leaves and stems. The proportion of the applied quantity of poisonous constituent retained by the plant decreases as the volume of spray is increased, and increases as the concentration of the spray solution is increased. Leaf sprays will therefore be most effective if the minimum volume of solution required for coverage is used, and the effective dosage of the chemical obtained by adjusting the concentration.A few indices sometimes used for estimating the efficacy of herbicides were compared with the mortality criterion. The number of leaves left, and the height of the living plants after treatment are of little or no value for estimating the effect of the chemical. The weight per unit area of the living plants remaining after treatment may be of some value.