The fact that the aqueous extract of 500 grams of the fresh _Astragalus mollissimus_, or of 200 grams (in some cases 100 grams) of the dried plant, when fed by mouth, would regularly kill a rabbit weighing about 907 grams, with certain definite clinical symptoms and pathological lesions, was at first arbitrarily selected as our test to aid in the isolation of the active principle. Later the production of chronic symptoms by the aqueous extract or digestion of 200 grams of these dried plants given in doses of 100 grams each on two successive days was considered essential. Carnivora, such as dogs and cats, vomit so easily as to render them unsuitable for these investigations. The aqueous extract was distilled with and without steam, also after acidifying with sulphuric acid, and likewise after the addition of magnesium oxid, but in all cases the distillate was inactive. The concentrated aqueous extract was shaken by the Dragendorff method with petroleum ether, benzol, chloroform, ether, and amyl alcohol, both in alkaline and acid condition, but the shakings yielded no physiologically active body. Shakings by the Otto-Stas method also proved inactive. Lead acetate, lead subacetate, silver nitrate, mercuric chlorid, alcohol, phosphotungstic acid, trichloracetic acid, ammonium hydrate, sodium carbonate, sodium hydrate, Mayer's solution, uranyl acetate, silver oxid, and barium carbonate also failed to remove the active constituent. They gave heavy precipitates in all cases, but these proved inactive. Hydrocyanic acid was sought for with negative results. The pathological lesions in the very acute cases suggested in some respects oxalic acid, a saponin, a metal, or perhaps a toxalbumin as the active principle, but none of the precipitants for saponins, such as lead and copper, or the magnesium oxid method yielded a body which was active. Proteids were excluded by the fact that the various proteid precipitants--alcohol, trichloracetic acid, lead subacetate, mercuric sulphate or chlorid, and salting out with ammonium sulphate and sodium chlorid (complete saturation and half saturation)--failed to give an active precipitate. Glucosidal or alkaloidal bodies were also excluded. On dialysing for twenty-four hours, some of the active principle went into the dialysate and some remained in the dialyser. Ether yielded a precipitate from alcoholic solution which failed to kill. The possibility of the activity of the plants being due to its normal acidity was excluded by neutralizing the extract with sodium hydrate and precipitating the salts with alcohol. The filtrate proved active after removing the alcohol.