36 
CONTRIBUTIONS TO CHEMISTRY AND MINERALOGY. [bull. 167. 
In the analyses by Jannasch aud Kalb the following examples are 
very striking: 
Si0 2 . 
B 2 3 
found. 
B 2 3 
calculated. 
Difference. 
Snarun 
35.64 
34.88 
35.50 
37.05 
9.93 
8.94 
8.34 
9.09 
10.40 
10.17 
10.35 
10.81 
—1.47 
—1.23 
—2.01 
— 1. 72 
Mursinka 
Buckworth 
Brazil 
It would hardly be wise to dismiss these variations as due to errors, 
especially when the summation of the analyses is good and the analysts 
are known to be trustworthy. Such errors on the part of either Jan- 
nasch or Riggs would be almost incredible, and I am therefore inclined 
to believe that the analyses are good, and that we should seek a cause 
for the variations. In my scheme of formulation the bivalent group of 
atoms = Al — B0 2 occurs. Replace this in part by the similar groups 
= Al — OH and = Al — F, and the variations are accounted for. This 
supposition satisfies the analyses completely, and covers the ground. It 
is in accord with all the evidence, even though its validity may not be 
definitely proved. iiy its application to the discussion of the analyses 
the divergencies between the calculated composition and the composi- 
tion as found can be notably diminished. 
But although the formulae which I have adopted serve to express the 
composition of all tourmalines, they £ J ill leave room for alternatives. 
Penfield and Foote, as well as myself, assume that tourmaline is a mixed 
salt containing distinct boric and silicic radicles. Future investigation 
may prove that it is really derived from a complex boro silicic acid, as 
yet unknown; and the same conception maybe true of other species, 
such as axinite, datolite, danburite, cappelenite, etc. A series of boro- 
silicic acids is theoretically conceivable ; and until this question has 
been considered, the constitution of all the minerals above mentioned 
must be regarded as unsettled. 
