Evolution and Devolution of the Elements. 45 
will be seen that we have assigned positions between cerium 
and tantalum to those which are known definitely. According 
to our theory, their resemblance is due to a special charac- 
teristic trivalent ring, which largely masks any other 
properties resulting from internal structure. Moreover, 
since so little is known of them chemically, it is unsafe to 
say whether we are justified or not. There are, however, 
three of them about which a little more is known, viz. praseo- 
dymium, neodymium, and samarium. Emerson Reynolds * 
has given several good reasons for supposing them to belong- 
to the iron group. One of these is the high paramagnetic 
susceptibility of their salts. Now we have already stated 
that the iron group is especially characterized by the mag- 
netic properties of its first three elements, which arise from 
the presence of an abnormal ring. As evolution proceeds, 
this ring is not destroyed, but still exists inside the atom, 
and hence by our theory we should expect the derivatives 
of the first three elements to exhibit this particular property, 
but to a less degree. So, if it is granted that these three 
elements may be placed among the platinum metals from 
which they at first sight differ so greatly, it is to be hoped 
that no objection will be taken to the more or less temporary 
positions we have assigned to the others. 
We have now discussed our theory of the method of growth 
of the atoms, and if our surmise is correct, would it not be 
reasonable to suppose that the table of atomic weights as 
rearranged would exhibit some quantitative relations between 
the weights of the atoms of the elements ? If the atomic 
weights of the elements of group I. are written down in order, 
and each subtracted from the one that comes after it, the 
following numbers result : — 
6-02 16-02 16-01 46-3 47*4, 
and these numbers may be regarded as the weights of the 
outer rings of the elements from lithium to caesium. 
Examining the other groups, it is found that in general 
the differences are again 6, 16, and some number between 40 
and 50 ; but whereas all the numbers belonging to the first 
class are very nearly 6 or 16, and that though the other 
numbers vary considerably, they approximate to a mean value 
of 46, showing a definite repetition and relationship between 
them. If we again consider the numbers 6, 16, resulting 
from group I., we note that they may be written : 
(1 + 2 + 3), and (1 + 2 + 3) +(1 + 2 + 3 + 4) 
respectively. 
* Chem. Soc. Journ. vol. lxxxii. p. 619 (1902). 
