6 
Suppose we have a column of length l, containing q units 
of colouring matter per unit of length, the medium being 
perfectly transparent ; then the light transmitted will be 
ak ql . Now, suppose the colouring matter, instead of being 
uniformly diffused through the whole column, to be con- 
fined to an extremely small section at the bottom of length 
V, so that the length of the column above the section may 
be still taken as l. The intensity of the transmitted light 
will be the same in both cases; hence k ql =k q 'K 
If the medium be not transparent, we may now suppose 
the column of length l above the section containing the 
colouring matter to be occupied by some medium that ab- 
sorbs light. Let p be its coefficient of transmission. The 
light incident on the bottom of the column is of intensity 
qM v . After penetrating the column the intensity will be 
ak q ' l p l ; but by what precedes, q[l'=ql ; so that the intensity 
of the transmitted light corrected for absorption by the 
medium will be a(k q p)\ If, then, we have two cylinders 
containing q and c[ units of colouring matter per unit of 
length, and columns of liquid l and we shall have the 
relationship (7c q p) l =^(k q p)l'. From this equation we may 
determine p in terms of k and known quantities thus : 
q'V—ql 
P = ki-* 
In some experiments on the absorption of light by carbon 
diffusions I noticed that when one diffusion was much 
stronger than the other there was a slight departure from 
the simple rule of colorimetry which held in other cases • 
this I thought might be due to the absorption of the water 
employed. The probability that this is the cause would be 
increased if the value of p deduced from one experiment 
