On Tribasic Phosphoric Acid. 175 



made to decide this point. In such a case the oxygen surely 

 would be liberated by the phosphorus combining with metals, 

 and in oxydizing a phosphide more phosphoric acid would he 

 obtained from the same weight of combined phosphorus than 

 from phosphorus in its free state. Among all morals iron is 

 principally remarkable for the facility with which a constant 

 compound of it with phosphorus may lie obtained. After ob- 

 taining such a compound, and analyzing it in a very ingenious 

 way, Berzelins obtained as a result 122.8 for oxygen, which is so 

 near the one found with free phosphorus, as to decide this point 

 completely; namely, that phosphorus does not contain oxygen, 

 and secondly, that PO 5 presents an exception to the rule. After 

 saying this Berzelins tries to show that phosphorus may still be a 

 compound body, as is very probably the case with nitrogen ; to 

 decide these difficult points, however, the utmost care and criti- 

 cism have to be used, and one must not be prejudiced in favor 

 of the one idea or the other. Chemists, who are not much ac- 

 customed to accurate working, have commenced enriching our 

 knowledge concerning proportions by analyses undertaken to 

 prove speculative views, and have correspondingly corrected 

 their results. This is an easy but a dangerous course. 



P>. Berzelins determined the equivalent of phosphorus to be 

 31.21. 



In the year 1810, Berzelins published an answer to a com- 

 munication of Davy's, in which the latter tried toshowthat the 

 oxygen of the three acids of phosphorus is as 1:2:4; stating 

 that his analyses are not accurate enough, and that by a great 

 number of analyses of this kind he has found that the hot an- 

 alytical methods are those by which the una voidable errors 

 are not more than thousandths. He never passed that limit, 

 unless accidentally. An analysis which varies less than ,,',,, 

 is still good ; if, however, in Bpite of the usual care, the results 

 vary more than n 1 f(T , we can regard it only as an approxima- 

 tion. These views are even now fundamental with regard to 

 the estimation of an analysis. Aftertheyear L819 we find the 



