145 
strength, the value of the correction will probably vary. It 
will be better not to have the difference large, whether the 
discs be external or internal, for when the differences are 
large any errors in the determination of the lengths of the 
columns have a greater effect on the calculated result. I 
also tried to make a rough estimate of the length of the 
cylinder, which might be covered without any perceptible 
darkening of the disc. I found that a black cloth cover 
investing the cylinder might be drawn down until it was 
about 3 5 or 32 cm. from the disc; this would vary with 
the dimensions of the window and the relations of the cylin- 
der to it. Also the length given is less than the value of 
x previously deduced ; but it ought to be so, for the light 
from a vertical window to illuminate a horizontal disc must 
pass obliquely through the solution. It would a]so follow 
that parts of the disc more remote from the window are 
darker than parts nearer, hence if the cylinders are of 
moderate radius, either the discs should be small and should 
be kept with their centres moving along the axes of the 
cylinders, or in the case of larger discs the determination of 
colour should be made by a comparison of similar parts of 
the discs. In the estimation of colour it is also not a matter 
of indifference, when we are given any tint as a limit, how 
we approach that limit. Suppose we have two cylinders, a 
and j3, full of coloured liquid, that in a being the darker. 
Now pour out from a until the colour seems the same as in 
(3 ; before reaching the theoretical division sight will fail to 
discern any difference of colour. Now if we proceed 
cautiously, as we approach the limit there will be a natural 
hesitation and tendency to stop, and it seems likely that in 
most cases in obedience to that feeling we shall stop with a 
column a little too long. Now suppose we start with the 
cylinder a empty, and pour fluid into it; as we again 
approach the limit cautiously we shall again have a tendency 
to stop, and inasmuch as before reaching the limit we pass 
