222 Dr Fyfe on the Illuminating Power of Coal-Gas, &c. 
or of heavy hydro-carbons, while the proportion of the other 
inflammable gases varies, which, though they consume oxygen, 
give out little light during their combustion, and we shall find 
that the amount of oxygen required gives no indication what- 
ever of the illuminating power. 
Thus, suppose the composition to be 
Olefiant, . : : : 13 13 13 
Carburetted hydrogen, . + 100 65 51 
Carbonic oxide, = ; 4 14 8 
Hydrogen, . : ; Sie 8 28 
100 100 #100 
the first would require 207, the second 180, the third 159, of 
oxygen, yet the illuminating power would be nearly the same 
in all. Supposing the heavy hydro-carbons to vary, and even 
to become considerable, yet the quantity of oxygen may not be 
in proportion, owing to the hydrogen and carbonic oxide, which 
require only half of their bulk of that gas for combustion. The 
mode of ascertaining the illuminating power by the shadow is 
one in which we may place the utmost reliance, provided we 
burn the gases with the same kind of burners, and pay parti- 
cular attention to the circumstances affecting the appearance 
of the shadow; for it is well known that the colour of the 
shadow varies even from the same gas, when the flames from 
different burners are contrasted ; besides, the reflection of light 
from surrounding objects will also occasion a difference. Great 
care is therefore necessary when conducting the trials in this 
way ; and it requires nicely adjusted metres, and a regular 
pressure, so that the consumpt shall not vary during the per- 
formance of the experiment. 
The other method which I formerly recommended is not 
liable to these fallacies. In the paper to which I have already 
alluded the results of numerous trials are given, in which the 
illuminating power, as shewn by the chlorine test, very nearly 
agrees with those indicated by the photometric process ; and 
these experiments were performed with every possible atten- 
tion to the circumstances likely to affect the results, so far as 
they were then known. In a paper subsequently published by 
