Vol. 58.] THE MINERAL ANALYSIS OF ROCKS. ign 
were due to the fact that the Author had had inaccurate mineral- 
analyses to work from. 
Mr. Hotranp said that he had frequently made a similar use of 
heavy liquids for the quantitative estimation of special constituents. 
in mineral mixtures like anhydrite and gypsum, quartz and magne- 
tite, and other cases in which there was a sufficient disparity of 
specific gravities to reduce and eclipse the experimental error. 
But he had failed to make a successful application of this method to 
ordinary crystalline rocks, and did not consider that it could ever 
replace, for the purposes of the practical petrologist, the recognized 
bulk-analysis by chemical methods. With rock-constituents con- 
taining microscopic inclusions of foreign matter, varying in degree 
of alteration, intimately intergrown on a minute scale (like the 
various types of pyroxenes and the microperthitic felspars), or 
subject to zonal variation by isomorphous replacement, it would be 
impossible to obtain a constant of error for application to the results 
of separation by heavy liquids. With well-crystallized, simple rocks, 
like some granites and gabbros, approximate agreement between 
this and the direct chemical method might be possible. But in the 
case of such rocks, whose minerals have been separated and analysed, 
there is in general no call for any process that does not surpass the 
chemical method in precision; while to new occurrences of rocks, 
whose constituents have not been so examined, the process is,. 
ex hypothesi, inapplicable. Even in the illustrative examples worked 
out by the Author, there are serious departures from the actual 
chemical analyses, especially in connexion with the lime and 
magnesia, indicating only partial success in discriminating between 
the rhombic and the monoclinic pyroxenes. Errors of a similar 
nature must always occur with couples or trios of minerals which 
are totally distinctin chemical composition, and yet exhibit differences 
of specific gravity less than their own internal variations. One has 
only to recall cases like the spinels and garnets, soda-pyroxenes and 
sphene, epidote and hypersthene, sillimanite and hornblende, or 
partly altered olivines with any but the densest of these, to show how 
frequently it must be impossible to make separations clean enough 
for chemical calculations. From the standpoint of the practical 
petrologist, therefore, the process is capable of very limited applica- 
tion, and could never be relied on as a general substitute for the 
ordinary laborious method of chemical analysis. Nevertheless, the 
quantitative estimation of rock-constituents might with advantage 
be practised more frequently, and for facilities to this end petro- 
logists would ever be indebted to the Author for his simple and 
ingenious diffusion-column. 
The Rey. J. F. Brake pointed out that the separation of minerals 
by means of heavy liquids was in some cases interfered with, 
theoretically at least, by the surface-tension between the mineral 
particles and the liquid. If the material had to be ground fine, so as 
to ensure that each particle contained only one mineral, its arrange- 
ment would depend not only on the specific gravity but also on the 
