The following experiments were then made and they seem to give definite evidence not only of the existence of an error but also of its magnitude. Carbonate of cadmium was made by dissolving pure cadmium in nitric acid, adding an excess of ammonia and a small quantity of ammonium carbonate. After standing for some time the cadmium carbonate was filtered off and rejected. The filtrate was treated with an excess of ammonium carbonate and the precipitated cadmium carbonate allowed to digest in it for some time. After washing by decantation several times the carbonate was transferred to a funnel containing a porcelain filter-plate, covered with a piece of ashless filter paper of slightly larger diameter, and washed thoroughly. with water. It was then transferred to a platinum dish, care being taken to avoid contamination with filter paper and heated gently to convert it into oxide. The resulting oxide was powdered in an agate mortar, returned to the platinum dish and heated to incipient whiteness for seven hours in a muffle furnace. The temperature must not be too high, otherwise the oxide will distill readily leaving no residue. The oxide is slightly volatile at good red heat as was observed in trying to make a determinant at this temperature by the oxide method. A weighed portion of the oxide which had been prepared from the carbonate in the manner described was dissolved in a weighed porcelain crucible and the resulting nitrate converted into the oxide again by heat just as in the oxide method. This constitutes experiment I. Experiments two and three were made in exactly the same way except that a platinum crucible was used instead of a porcelain one. The results are: Initial Wt. Final Wt. Gain Corresponding Error in At. Wt. I. 2.95469 2.95650 .00081 −.24 II. 2.67717 2.67835 .00117 −.39 III. 3.00295 3.00422 .00127 −.38