386 PROCEEDINGS OF SECTION B. 
A special instance of this occurred to me lately with some 
phonolite rocks from the Western District of Victoria, in the 
felspathic matrix of which tolerably large tabular crystals of 
sanidine were observed. About twenty grains weight of the 
crystals was separated from the rock and analysed. As soda and 
potash exist in varying proportions in sanidine, it was necessary 
to be very accurate-in estimating them, and I attempted to 
control the gravimetric determination by a volumetric one. These 
two methods gave, however, such divergent results that I made 
special experiments to ascertain which of them was to be relied 
on for the estimation of small quantities of the alkalies. 
The volumetric method used was that of Fr. Mohr, in which 
the chlorine is determined by a solution of nitrate of silver of 
known strength, with potassium chromate to indicate the end 
of this reaction; and the gravimetric, the ordinary one with 
bichloride of platinum. : 
Nearly equivalent quantities of chemically pure sodium chloride 
and potassium chloride were taken, the actual weights being 
4:19 grains and 4:32 grains respectively. These being mixed 
their solution was divided into exactly equal portions, so that 
each half contained 2-095 grains of the sodium, and 2-16 grains 
of the potassium chloride, or 4°255 grains of the mixed salts. The 
actual weight of chlorine required by theory in this mixture of 
the two salts is 2°296 grains, or 53°96 per cent., and the amount 
found volumetrically was 2°31475 grains, or 54:4 per cent., the 
difference being thus—0-44 per cent., which falls within the limits 
of reasonable error, and is as near as volumetric determinations can 
be expected to reach. The formula for calculating from the 
chlorine found the respective weights of the salts present was the 
usual one, which gives an answer within one per cent. of the 
truth, when the amount of chlorine is exactly ascertained. I say 
exactly, because the slightest error in this becomes much magnified 
by the multiples used for finding the relative weights of sodium 
chloride and potassium chloride in the mixture of the chlorides. 
Thus, although the initial error in my experiment was only 0:44 
per cent. of chlorine, yet, by applying the formula, I obtained 
6:1 per cent. more of sodium chloride and six per cent. /ess of 
potassium chloride than the solution really contained. 
The remaining half was treated with platinum bichloride, the 
resulting precipitate washed with spirits of wine, and the double 
salt of potassium and platinum left in the dish weighed. 
Calculated to potassium chloride, its weight represented 2°1576 
grains, or 99°88 per cent. of the amount taken, viz., 2°16 grains. 
The sodium chloride was calculated from the difference. 
Other experiments made have given the same results, z.e., the 
gravimetric estimation accurate and the volumetric inaccurate 
The amounts of the two chlorides being nearly equal in these test 
experiments ; the volumetric method was applied under the most 
