142 ABSOLUTE ATOMIC WEIGHT. 



forming cross-ties, giving strength to the entire system here 

 erected on a field cleared of the errors of Stas. 



The chemical process selected is the warming, and sub- 

 sequent distillation of anhydrous borax with methyl alcohol 

 and muriatic acid, leaving dry sodium chloride as residue. 

 See pp. 52-53. 



Both the substance and the final product are fixed, 

 accurately weighable solids. 



The quantitative chemical reaction is expressed in the 

 following atomic ratio: 

 2 Na Cl : Na2 OT Bo* zz: 117 : 202:^:0.57 921. Chg. 115 low. 



Two series of determinations were made. In the first 

 series, flasks of not very resistant glass were used and found 

 to be attacked sufficiently to vitiate the perfect accuracy of 

 the results, giving as mean analytical ratio 0.57 948 which 

 is 27 high. 



This result should, therefore, not be used. It has prop- 

 erly been rejected, for cause stated as above, by the experi- 

 menters themselves. Of course our own Chief Chemist put 

 this condemned morsel as a tit-bit into his fragrant olla 

 podrida. 



For the second series, flasks of very hard glass (combus- 

 tion tubing) were used and the error, though not absolutely 

 avoided, was reduced to a very minute amount. 



Now the sign of this error is perfectly known ; it is an 

 increase of the ratio which, therefore, should come out 

 high, if no other constant error affects the process chosen. 



On account of the extraordinary importance of this 

 second series of determinations, I will transcribe the weigh- 

 ings from the journal referred to. 



I shall, however, omit the last two decimals of the 

 weights given in the journal. These are beyond the range 

 of positive determination, and therefore not experimental 

 data. They have, as a matter of fact, completely disappeared 

 in all reductions made by the authors themselves. I speci- 

 ally call attention to this fact. 



Probably it is the recommendation of Clarke, quoted 

 under thallium that induced Ramsay to copy the model of 



