As we started with cadmium oxide, and, after passing to the nitrate, converted it back into the oxide, the weight should remain unchanged if the method is correct. However, this is not the case, but a large increase in weight takes place. The increase is larger in a platinum crucible than in a porcelain one, which accounts for the fact that a lower value for the atomic weight is found by the oxide method when they are used. The use of a porcelain crucible therefore diminishes the error, but does not eliminate it. The explanation of this has already been given. The oxides obtained in these three experiments were tested for occluded gases in the manner already described, but only small amounts were found. Both of those made in platinum crucibles were tested for nitrate of cadmium with brucine and sulphuric acid with negative results. To show that the impurity was not converted into an ammonium salt when the oxide was dissolved in hydrochloric acid, a slight excess of caustic potash was added to the solution, the precipitate allowed to subside and the clean, supernatant liquid tested for ammonia with Nessler’s reagent. No ammonia was found. In order to make these experiments as severe a test as possible, a somewhat higher temperature was employed than had in the five experiments described under the oxide method. This was accomplished by boring out the stopcocks of the blast lamp so that a larger supply of gas was furnished. The two oxides in the platinum crucibles seemed to be constant in weight, but that in the porcelain crucible seemed to lose in weight slowly. The weight given was taken after four hours blasting, which is longer and at a higher temperature than was used in any of the five determinations made by the oxide method. If the cadmium oxide prepared from the carbonate retained any carbon dioxide, it would lose weight in being dissolved and reconverted into oxide. The above experiments therefore seem to furnish very strong evidence that there is an error of at least −.24 unit in the oxide method when porcelain crucibles are used and −.39 of a unit when platinum ones are employed. if .24 of a unit is added to 112.07 the result obtained when porcelain crucibles are used we get 112.31 and adding .39 to 111.87 gives 112.26. Considering the small number of experiments made, the fact that they were made in such a way as to give a low value (numerically) for the error rather than a high one, and also that the error is probably variable to some extent, especially when porcelain crucibles are used, the corrected results agree as closely with 112.38, the average of the chloride, bromide and sulphate (synthetical) methods as could be expected. It must also be borne in mind that 112.38 is only to be regarded as an approximation to the atomic weight of cadmium. The increase in weight observed in converting the nitrate back into oxide might also be explained by assuming that the cadmium oxide used in the beginning of the experiments was richer in