B.--On the Separation. First, from its soluble simple combinations with the acids or bases containing no iron or cobalt; if they are present, it is treated as is noted later. If sulphuric acid is present it must be separated by treating the solution of the compound with barium chloride and filtering. A nearly neutral solution is prepared in water or hydrochloric acid and placed in a flask. Here it is treated with chlorine by passing a current of that gas through it as long as it causes a precipitate and for some time afterward. It is then discontinued, the mixture allowed to deposit for a few moments, and about two-thirds of the supernatant solution decanted; it is mixed with some more water, and these decantations repeated until they pass away without reaction, or by filtering it and washing on the filter; it is then dissolved in hot hydrochloric acid, this nearly neutralized, a solution of sesquichloride of iron is added, and again treated with an excess of chlorine. After washing it is transferred to the flasks of the apparatus mentioned in the first part of this paper, and estimated. Myself and several others have found this always to be a true MnO_{2}, and not a varying mixture of protosesquioxide and binoxide, and will thus yield accurate results. This reprecipitation may sometimes be dispensed with by adding the iron salt before the first precipitation, but it of course depends upon the other elements present. From Compounds containing Cobalt, Cobalt and Nickel, Iron and group III., together or with other elements.--Group III. and sesqui. iron are separated by agitation with baryta carbonate, some chloride of ammonia being added to prevent nickel and cobalt precipitation traces, and filtering. If cobalt is present we treat this filtrate with nitrite of potassa, etc., to separate it (that is, if it and nickel are to be separated and estimated in the same sample; but if they are to be estimated as one, or not separated, the treatment with nitrite, etc., is not used). The filtrate from this last is directly treated with chlorine. If nickel and cobalt are not to be estimated in this sample, the solution, as chlorides, is mixed with some chloride of ammonium and ammonia, then with a fair excess of phosphoric acid, a sufficient quantity more of ammonia to render the mixture alkaline. The precipitate formed is transferred to the filter and well washed with water containing NH_{3}Cl and NH_{4}O, then dissolved in hydrochloric acid and reprecipitated with ammonia, filtering and washing as before. It is again dissolved in HCl and titrated with uranium solution, or decomposed by tin, as noted below, and the manganese precipitated as binoxide with chlorine, and determined. The latter method is hardly practicable, and I never have time to use it, as the titration and all together yields a value of 99.80 in most cases, if accurately executed.