139 



1st Series. 



These numbers indicate that even the natural oscillations of atmo- 

 spheric pressure must produce a considerable variation in the amount 

 of light emitted by gas-flames, and it was therefore important to 

 determine, by a special series of observations, this variation in lu- 

 minosity within, or nearly within, the usual fluctuations of the 

 barometrical column. In order to attain greater delicacy in the 

 pressure readings in these experiments, a water-gauge was used, 

 but its indications are translated into inches of mercury in the 

 following tabulated results, each of which represents, as before, the 

 average of twenty observations. 



3rd Series. 



It is thus evident that the combustion of an amount of gas which 

 would give a light equal to 100 candles when the barometer stands 

 at 31 inches, would give a light equal to only 84 - 4 candles if the 

 barometer fell to 28 inches. 



