saturated with vapor at the temperature of the experiment. During the course of the analysis, the temperature of the measuring flask, which is almost entirely surrounded with water, does not vary sensibly, but in a series of experiments which are executed at different times, the temperature of the measuring apparatus, which is that of the ambient air, may change much. It may oscillate between 10° to 25°, and exceptionally between 0° and 30°, whence there are notable variations in the tension of the vapor of the gas measured. If it should be desired to calculate to 100 parts of dry gas the observations made at 30° upon 100 parts of saturated gas, it would be necessary to increase the percentage of carbon dioxid by about ¹⁄₂₅ of its value. It is noticed that with the apparatus described above, the gas upon which the estimation is really conducted comprises not only that which the measuring apparatus contains from E to C before the absorption of the carbon dioxid, but also the small quantity which remains in the capillary tube KME at the moment when closing the pinch-cock M, after the second rinsing, the gas from the soil is aspired into EAC. On the other hand, there is left in the same tube KME, when the final reading is made, some gas which belongs to that which has been measured at the end. These two small gaseous portions which we consider in the tube KME to be sensibly equal, do not contain any carbon dioxid and may be left out of consideration. That is why the volume of the measuring apparatus is limited to E and the graduation of the tube BC is in hundredths of the volume comprised from E to C. In reality the two portions are not absolutely equal because the two successive levels of the potash solution, which limit them in the tube K, are not absolutely identical. These two levels can differ in such a manner as to correspond to a volume of about ¹⁄₁₀₀₀ of the measuring apparatus. Thus the estimation is really made upon a volume of gas which may be greater or less by ¹⁄₁₀₀₀ than the volume of EAC; whence there might result an error of ¹⁄₁₀₀₀ in the estimation of the carbon dioxid, an error which is wholly negligible.