OF ARTS AND SCIENCES. 69 



Kissel's method of compensating for enclosed material by excessive 

 washing, the amount of washing must necessarily be proportioned 

 both to the weight of phosphoric acid present and to the excess of the 

 precipitant, — elements which introduce a great deal of uncertainty, — 

 it would appear that if magnesic sulphate mixture is to be the pre- 

 cipitant the method of precipitating twice is the only one which can 

 be relied upon on all occasions to give tolerably correct results. 



The results obtained in experiments (27) to (66) by precipitating 

 with magnesic chloride mixture are different as the method of treat- 

 ment varies. It will be noticed that in experiments otherwise essen- 

 tially similar the variations in the amount of the precipitant produce 

 differences in the results. Thus, in experiments (27) to (43), in 

 which 10 cm. 8 or 12 cm. 8 of the precipitant were used, the error 

 varies from —0.0008 gr. to -4-0.0012 gr., or from —0.42 to -f-0.64 

 per cent., the mean error being nearly ; while in the experiments 

 (48) to (53) in which 40 cm. 8 of the precipitant were used, the error 

 varies from -{-0.0011 gr. to -4-0.0016 gr., or from -4-0.60 to -j-0.87 

 per cent., the mean error being about -4-0.0013 gr., or -4-0.71 per 

 cent. It will he seen also that while it makes no apparent difference 

 whether the precipitant is added to the phosphate directly, as in (27) 

 and (28), or after treatment with hydrochloric acid, as in (29) to 

 (39), the dissolving of a precipitate, by adding hydrochloric acid to 

 the supernatant liquid, and reprecipitation with ammonia tend 

 to reduce the results below those of experiments in which this 

 has not been done, but which are similar in other respects. This 

 appears very distinctly in a comparison of experiments (40) and (41) 

 with (42) and (43), where the mean error is reduced from about 

 -4-0. 00 10 gr. upon 0.1875 gr. of phosphoric oxide, or from -4-0.56 

 per cent., to ; or of experiments (44) and (45) with (46) and (47), 

 where the mean error is reduced from -j-0.0022 g r- 1 about -)-0.0003 

 gr. upon 0.1885 gr. of phosphoric oxide, or from -(-1.17 to -j-0.18 

 per cent. 



It appears, further, from a comparison of experiments (54) to (58) 

 with experiments (48) to (53), that results got by precipitating 

 solutions containing either no free ammonia, as in (56), (57) and 

 (58), or only a small amount, as in (54) and (55), by a magnesic 

 chloride mixture deprived of ammonia are more nearly correct than 

 those got by precipitating under essentially similar conditions with the 

 ammoniacal magnesic chloride mixture as in (48) to (53). Thus, in 

 (48) to (53) the error varies from -4-0.0011 gr. to -(-0.0016 gr. upon 

 0.1831 gr. of phosphoric oxide, or from -j-0.60 to -(-0.87 per cent., 



