70 JOURNAIv O^ THE 



namel3% that of taking* the work of different authors, 

 and critically averag^ing- it, must be rejected, for this 

 would never secure concordant results as they would 

 have to depend upon the judg'ment of the critic and 

 calculator. All of this back work must be wiped out 

 of existence, except for historical use, and we must 

 beg^in anew with every conceivable refinement of 

 method and apparatus, perhaps devoting-, as has been 

 sug-g^ested, some central endowed laboratory to that 

 work and that alone. 



Calculations of "probable errors " have g-iven a seem- 

 ing* accuracy to many atomic weig^ht determinations. 

 In this chemists have followed the lead of Stas. It 

 seems to me that this is ver}' misleading-, these calcula- 

 tions often being- made upon small series in which the 

 possible error, as shown by the variation in individual 

 experiments, is ten times that shown by the calcula- 

 tion. In the proposed new determinations this method 

 of calculation should only be allowed in the case of 

 several hundred closely ag-reeing- determinations. 



Of course, it might as well be confessed that abso- 

 lute accuracy is not to be even hoped for. The best 

 methods and appliances which can be devised or man- 

 ufactured will always be imperfect and there is besides 

 the personal error of the observer to be allowed for. 

 To what extent shall we demand accuracy, then ? 

 Where shall the line be drawn ? Is it to be at the first 

 decimal place or the second ? It seems useless or hope- 

 less to speak of the third. There can scarcely be 

 said to be an atomic weigfht at present known correctly 

 to the first decimal place. Take the numerous deter- 

 minations for oxyg-en, exceeding-ly modern and excel- 

 lently well carried out, and see how they vary between 

 16.0, 15.9 and 15.8 and look at the original series from 



