In the first five determinations, the analytical operations were conducted as nearly as possible alike, but the preparation of the portions of cadmium chloride taken for analysis was varied very much as will be seen by referring back to this part of this paper. The results do not vary more than ±0.015 from their average. This is very strong evidence of the purity of the chloride used for, if it contained any impurity, we should have expected to vary the amount in the different portions. After this, attention was paid especially to the analytical process, for it was thought that there probably was some serious error in the method, the result being higher than any that had previously been obtained, if we exclude Dumas’ first series which he himself did not accept. The conditions were varied in many ways to see how much the result could be influenced, but under no conditions were results as low as Huntington’s average (112.24) obtained. A number of errors were found in the method during the work, but they seem to neutralize each other to a great extent. The more important ones will now be given. Nearly every filtrate including the corresponding wash water was examined for chlorine after the silver and cadmium had been precipitated by hydrogen sulphide. The excess of hydrogen sulphide was expelled by boiling, after the addition of some nitric acid. In two cases an inverted condenser was used. On adding silver nitrate a precipitate was always obtained showing the presence of chlorine. Care was always taken to filter off sulphur formed by the oxidation of hydrogen sulphide, before adding the silver nitrate. The precipitate was never very heavy, and was not estimated quantitatively. It is evident that cadmium nitrate exerts a solvent action on silver chloride. In some cases a very large excess of silver nitrate was added but it did not change the results markedly. Silver nitrate itself dissolved silver chloride to some extent. The increase in insolubility, if any, on adding an excess of silver nitrate is probably counterbalanced by the increased error due to occlusion of nitrates in the silver chloride. Stas (Aronstein’s Trans. p. 156) says it is impossible to contract silver chloride or bromide in a solution containing salts without there being occlusion and that the precipitate can only be freed from them by dividing up the contracted mass by shaking with pure water. This was not done here owing to the solubility of silver chloride in pure water, and the complications introduced in the analytical part. The occlusion of nitrates by the silver chloride would lower the atomic weight found. The silver chloride obtained always darkened on heating and contained cadmium, as was shown in the following manner: The lump of silver chloride was attached to the negative pole of a cell and electrolyzed in a bath containing dilute sulphuric acid. The resulting metal was then dissolved in nitric acid and the silver precipitated by adding hydrochloric acid.