OF ARTS AND SCIENCES. 71 



The differences between results got by using a magnesic sulphate 

 mixture and those obtained with a magnesic chloride mixture may be, 

 perhaps, partly explained by the supposition that magnesic sulphate 

 resists exclusion during the crystallization of the ammonio-magne^ic 

 phosphate more effectively than magnesic chloride, and partly by the 

 fact that magnesic sulphate enclosed in a precipitate would probably 

 not change materially in composition during an ignition over a Bun- 

 sen burner in a crucible standing on platinum foil, while magnesic 

 chloride ignited in presence of aqueous vapor under like circumstances 

 would be converted into magnesic oxide whose molecular weight is 

 but one-third of that of the sulphate. It is difficult, at all events, 

 to conceive how the contamination of the precipitate can be other 

 than mechanical ; for the trimagnesic phosphate could only be pro- 

 duced by a reaction between the ammonio-magnesic phosphate and 

 magnesic sulphate, — which is quite improbable, at least in the cold, — 

 and the formation of a magnesic hydrate or basic sulphate would 

 seem to be altogether unlikely under the circumstances. 



In the experiments upon the phosphotungstates the mean error of 

 the method involving a single precipitation with magnesic sulphate 

 mixture (experiments 67 to 70) was about -(-0.0021 gr. upon 0.0265 

 gr. and with magnesic chloride mixture (experiments 80 to 84) about 

 -j-0.0024 gr., upon 0.0266 gr. or 0.0269 gr., of phosphoric oxide, or 

 about -(-8.44 per cent. The mean error of the method of double 

 precipitation upon nearly the same amounts of phosphoric oxide, the 

 precipitant alone being used to wash the first precipitate and no citrate 

 being present, was, with magnesic sulphate mixture (experiments 75 

 to 79) about -{-0.0004 gr., or -{-1.51 per cent., and with magnesic 

 chloride mixture (experiments 89 to 91) -f-0-0002 g r-) or _|_o.74 per 

 cent. ; the mean error when the first precipitate by magnesic sulphate 

 mixture was washed with the precipitant and afterwards with ammonia 

 water (experiments 71 to 74) being about — 0.0015 gr., or — 5.66 

 percent.; and when the first precipitate was thrown dwn in pres- 

 ence of a citrate by magnesic chloride mixture and washed with the 

 precipitant (experiments 85 to 88), about — 0.0007 gr., or — 2.59 

 per cent. "When precipitation was effected with 1.5 cm. 3 of magnesic 

 chloride mixture, the same being added after precipitation (experi- 

 ments 92 to 95) the mean error upon 0.0269 gr. of phosphoric 

 oxide was -(-0.0004 gr. or -(-1.49 per cent. ; when by 2 cm. 8 (experi- 

 ments 96 and 97) about -(-0.0006 gr., or -(-2.23 per cent. ; when 

 by 5 cm. 3 (experiments 98 and 99) -fO.0011 gr., or -(-4.09 per 



