148 
result happened when I hung up yellow screens and tried 
to make determinations of colour behind them ; also when 
looking at light yellow external surfaces, differences in the 
lengths of the columns failed to give any difference in tint, 
although when looking at white external surfaces they did 
so. But in quantitative determinations of matter by colo- 
rimetry, the excellence of the results require sensible varia- 
tions in colour when we alter slightly the length of the 
column ; hence when the incident light is tinged with the 
colour we wish to determine the advantage of the method is 
diminished. Such a consequence may also be deduced from 
the formula which I obtained in my last paper. For sup- 
pose white light to consist of yellow, blue, aud red (as far 
as the reasoning is concerned we might have considered it 
also composed of green, red, and violet, as some physicists 
do). Let I denote the incident white light, and B, Y, It the 
intensities of blue, yellow, and red necessary to produce 
white light, so that we may write : 
I = B + Y + E. 
Let there be two solutions containing q and q' of yellow 
colouring matter, and t and t' the corresponding lengths of 
columns; then the intensity of the light transmitted through 
one cylinder will be 
(1 - mqt ) Y + (1 - m x qt)R + (1 - m x qt) B, 
m denoting the amount of yellow light absorbed by a unit 
layer, and m x the amounts of red and blue absorbed by a 
unit layer. Also the light transmitted by the other cylinder 
will be 
(1 - mqt 1 ) Y + (1 - miqt'JR + (1 - mqt') B. 
Since both cylinders are of the same colour these expres- 
sions will be equal, m will be less than m x because the 
transmitted light is yellow ; let m = m x - fi. Then we shall 
have 
1(1 - m x qt) + fiqtY - 1(1 - m x qi) + fiqt! Y) (A) 
The expression on the right hand denoting the light trans- 
mitted through one cylinder, and the expression on the left 
