672 TRANS. ST. LOUIS ACAD. SCIENCE. 



atomic weight would be reduced by an amount proportional to 

 the quantity of the unknown elements and inversely proportional 

 to their atomic weights. But the effect of each separately, and of 

 all together, would be to redtice the atomic weight of I. 



Consequently we find the atomic weight of this element, as de- 

 termined by experiment, is much too low. 



In the case of Br, however, ^^, y.-^ and j'g would tend to reduce, 

 while y j^ and j ^ would tend to raise the atomic weight. If all 

 five were present in about equal proportions, the resultant would 

 still be too low ; not, however, so much as in the case of I ; and 

 this we find to be the fact, Br being three and I four units too low. 



If these bodies really exist in Br and T, it is reasonable to sus- 

 pect that C/ may also be contaminated with some of them in very 

 small, but nearly constant, quantities. The entire C/ supply of the 

 world comes practically from the sea, or from beds of rock-salt 

 derived from the sea. On this account, if any of these unknown 

 elements exist in any C/, they probably exist in all C/ in about 

 the same proportion. 



It will be seen at once that all except y^ would raise the appa- 

 rent atomic weight of C/. This gives a very satisfactory explana- 

 tion of the irregular atomic weight of C/, it being .37 too high. 



This .37 makes C/ the only exception to Prout's law, and to the 

 law of equicrescent valence between H and Car. 



In view of these facts, does it not seem possible and even pro- 

 bable that some of these elements exist in C/, Br, and I ; but are 

 so closely related to them that it will require more delicate tests 

 to distinguish them than any we have yet been able to apply? 



The same argument will apply in exactly the same way to the 

 K group, and with even greater force. 



And may we not look upon Saxony lepidolite or rock-salt as a 

 key which only awaits the hand of a Bunsen or a Kirchoff" to un- 

 lock the whole mystery.? 



In this way I venture to explain the deviation of A5, S^, S«, 



C5, R(5, Y, and of Se, if the atomic weight of the last element 



must be considered less than 80. +1 



The broken curve shown on the plates passing through Ag , 

 —3 -1 



+ 2 +3 +* +5 +7 _ 



Qd , \n , S« , S<5 , and I , indicates a nearly constant negative 

 error of several units in the atomic weights of these elements. 



