32 THE ERRORS OF PRECISION. 



uncertainty of weighing the corresponding uncertainty in 

 the atomic weight is 0.055 or practically 0.06 units in Hg. 



But the greatest decimal given by Hardin is precisely 

 0.06. The determinations of Hardin are uncertain to this 

 extent. Hence these fractions of Hardin are not experi- 

 mentally determined. They represent nothing but his own 

 imagination. 



For the fifth determination the change per unit (in hun- 

 dredth mgr.) amounts to o.io; hence the uncertainty in the 

 value of Hg is 0.05. The fraction 0.03 of Hardin is within 

 this limit of uncertainty. His statement Hg = 200.03 * s n t 

 a statement of fact, is not warranted by his weighings. He 

 is drawing on his own imagination. 



For the fourth determination, in which the weight of 

 the oxide taken is stated to be 141.48 mgr. the change (per 

 unit in hundredth milligramme) is 0.20; hence the uncer- 

 tainty in the atomic weight Hg amounts to o.io. 



Now in this instance, Mr. Hardin gives Hg = 200.00; both 

 of these decimals are due to his own imagination, since his 

 weighings leave the atomic weight uncertain to the unit in 

 the first place. 



In other words, in the stated value Hg = 200.00 Mr. 

 Hardin assumes for his determinations an accuracy one 

 hundred times the actual precision of his own weighings! 



Such statements, pretending to be statements of experi- 

 mental facts, are a blot upon science and block the way of 

 the truth; for they are false in fact, merely imagination and 

 fancy. They are essentially fraudulent, for they do not 

 present fact as they pretend to do. 



VII. ERRORS DUE TO FALSE DATA. 



The second mathematical error universally committed by 

 the chemists of the present is in itself fatal to the produc- 

 tion of a true value of the atomic weight or even a correct 

 expression of the often admirable chemical laboratory work 

 done. 



We refer to the common habit of "adopting" some set 

 of values of atomic weights in the calculation required, 



