xvni THE CLASSIFICATION OF THE ELEMENTS 483 



The average of five experiments actually gave the ratio 8 : 

 i6'oi, agreeing closely with the modern ratio oxygen : sulphur 

 = i6:32'o7. Dumas (still using equivalents rather than 

 atomic weights) also gave 



Co : Ni : Sn= 29-5 : 29-5 : 59 = i : i : 2 and 



N : Fe : Cd = 14 : 28 : 56 = i : 2 : 4 ; 



these numbers may be compared with Crookes' suggestion 

 (Phil. Trans., 1908, A. 209, 44) that 



B : Sc : Y : Yb= u'o : 44*1 : 89*0 : 173*0 = i : 4 :8 : 16 

 No special importance is now attached to the fact that 

 the atomic weight tables do actually show some very close 

 multiples of the atomic weight of oxygen, or that this 

 number, as Newlands pointed out in 1864 (The Periodic 

 Law, p. 6) is even more common amongst the differences ; 

 thus, when O= 16, 



8 = 32-07 Ti = 48-i Br = 79-92 Mo = 96-0 

 8-0=16-07 K-Na=i6'io Na- Li= 16-06, etc. 



It is, however, remarkable that integral atomic weights are 

 much more common when O=i6 and H=i'oo8 than 

 when H=i and 0=15-88. 



Experiments to test Prout's hypothesis. Thomson, who 

 gave to Prout's paper the place of honour in his annual 

 review of the progress of Chemistry for 1815 (Thomson's 

 Annals of Philosophy, 1816, 7, 17) confirmed Prout's hypo- 

 thesis by a series of experiments carried out between 1819 

 and 1825 (An Attempt to Establish the First Principles of 

 Chemistry by Experiment, London, 1825). But he appears 

 to have had an unconscious bias in favour of those experi- 

 ments which gave " correct " or integral values for the atomic 

 weights. In the words of Berzelius : 



" He reduces all the numbers found by his predecessors to 

 the nearest multiple of the atomic weight of hydrogen, 

 calculates therefrom the atomic weights of their compounds, 

 and precipitates them in weighed quantities corresponding 

 with the corrected atomic weights, when they always 



I I 2 



