28 THE ERRORS OF PRECISION. 



number exactly; but the variations must lie on either side of 

 a horizontal line in a diagram of the kind here used. 



If a chemist engaged by a firm to analyze a limestone 

 should hand in a report showing clearly that by his method 

 of analysis he had found the percentage of lime to run up 

 (or down) with the amount of limestone operated upon by 

 him, the firm would most likely not entrust any further work 

 to this chemist. 



We regret to find by these symptoms, that Professor 

 Mallet has suffered from Morbus Stasii for many years. 

 This should be remembered before judging his work harshly. 



This is so much the more called for as we notice his case 

 complicated by the incipient stages of Furor Clarkii. 



Our Conclusion. 



Therefore, the only legitimate conclusion that can be 

 drawn from the results given by Professor Mallet and repre- 

 sented graphically to scale in our diagram, is that the 

 methods used are not chemically satisfactory, because they 

 give results showing the existence of errors varying with 

 the amount of matter operated upon. 



Instead of taking the mean of eaCh series as the true 

 value, the entire work should be discarded as tmreliable. 



We have purposely dwelt upon these details of the noted 

 determinations of the atomic weight of gold by Professor 

 Mallet because the recognized high value of the chemical 

 work shows strikingly the necessity of some critical exami- 

 nations of such work by methods not yet used in the 

 chemical laboratory, but which can be drawn from the 

 general science of quantity and form, that is, from mathe- 

 matics and common sense. 



VI. ERRORS IN PRECISION. 



Incredible as it may seem to any one not familiar with 

 the causes that have led to the existing practice of calcula- 

 tion among chemists engaged in the determination of the 

 atomic weights of the elements, we find, as a matter of fact, 

 that two errors of method are quite generally committed, 

 either one of which would suffice to vitiate the results pro- 



