174 PROF. W. J. SOLLAS ON A PROCESS FOR [May 1902, 
to state the result :—The ratio of the loss of orthoclase and horn- 
blende was found to be 22°58: 17-82, which is almost identicai 
with that obtained when the minerals were crushed together. On 
now examining the separated powders under the microscope, it was 
found that the orthoclase contained a notable admixture of horn- 
blende, while the precipitated hornblende was almost free from 
orthoclase. The change in the ratio is evidently due to imperfect 
separation ; but as the separation was performed very carefully, it 
is clear that, within the limits of practicable as opposed to possible 
analysis, this imperfection must always be reckoned upon. It 
will therefore become necessary te experiment on all the chief rock- 
forming minerals, and to determine once for all the constant of 
error to be applied in each case. 
So far, the proportion of loss has been determined in only a few 
cases. They are as follows :— 
Orthoclase : Hornblende = 22: 
Quartz : Hornblende = 19: 
Quartz : Olivine 
| 
bo 
b Lay 
CO 
=] 
— 
or 
oO 
On 
The next step was to ascertain how far the presence of flour in 
the separator would affect our results. Orthoclase and hornblende 
were powdered together, the mixture containing 61°65 per cent. of 
orthoclase and 38°35 per cent. of hornblende. Without any prelimi- 
nary washing this was introduced into the separator, and yielded 
59°65 per cent. of orthoclase and 40°35 per cent. of hornblende; or 
almost precisely the same relative amounts as were obtained from 
the washed material, which gave 59°71 per cent. of orthoclase and 
40:29 per cent. of hornblende. 
In all these cases, acetylene-bromide and benzene were used for 
the separating fluid, but it is doubtful whether the lower viscosity 
of a fluid produced by mixing methylene-iodide and carbon-tetra- 
chloride would greatly affect the results. The conclusion, however, 
is clear that a preliminary washing of the powdered rock may 
conduce to greater accuracy, provided that we have previously 
established the constant of error for each mineral constituent. Of 
course, in order to apply this we must always inform ourselves by 
weighing of the amount of material removed in the process of 
washing. The saving of time thus ensured is twofold. In the first 
place, separation can be produced more rapidly, and in the next the 
number of necessary weighings is diminished one-half, for in the 
case of grains of appreciable dimensions they may be turned out of 
the filter-paper into a capsule without leaving an appreciable 
quantity behind, and thus the labour of incinerating the filter and 
weighing the ash is avoided. 
It may be observed that the general tendency of the errors 
incidental to separation is to give an under-estimate for minerals of 
lower specific gravity, an over-estimate for those of higher specific 
