234 CLARKE'S CONFESSION. 



Other Determinations. 



Thorpe and Laurie, 1887, made, independently of Kriiss, 

 a research partly overlapping. They subject the same com- 

 pound to reduction by direct heat (ignition), which in itself 

 is better; the residue was weighed, and gold determined by 

 washing out the bromide, and weighing the residue. They 

 thus give up the original compound, on account of their 

 impossibility of drying it completely without beginning 

 decomposition, a difficulty apparently overcome by Kruss. 



The low Brs in Kriiss points to some loss in this direc- 

 tion. But surely, Thorpe and Laurie obtained very inaccu- 

 rate results from the residue. Instead of 0.60 406 they found 

 the mean 75 low; in this case, the Ka Br was determined by 

 difference only. 



The silver work of these analysts needs, of course, no 

 attention here. 



The determinations of Mallet, 1889, have been fuHy 

 exhibited in an earlier section of this work. See pp. 24-28. 



Clarke Condemns His Own Work. 



In this connection we must quote the following from 

 page ico of Clarke's Constants of 1897 (we insert letters for 

 reference) : 



a The former agreement between the several series of 

 ft gold values (a) was therefore only apparent (b), and we 

 " are now able to see (c) that concordance among deter- 

 " minations may be only coincidence (d) and no proof of 

 " accuracy (e)." 



Our references: 



(a) We did see no such agreement; see diagram, Plate I. 

 Clarke's formal statement of fact, is false. 



(b) Was not apparent, was none there at all. 



(c) Cannot see concordance of several series or their 

 individual determinations yet. 



(d) Everybody knew that long ago; but we must still 

 declare the absence of " concordance " in the results of 

 Mallet, as given by himself and represented on our diagram 

 to an exact scale. 



