of the Constituents of Water, &c. 205 
recipient attached to the above-mentioned tube, in order to allow 
us to examine its purity: in the subsequent ones, the aqueous 
vapour and the excess of the hydrogen passed through a long 
column of fused muriate of lime. It is easy to see of how much 
precision this mode of performing the experiment is capable. 
Hence the results obtained in the several trials differ but little 
from each other; and as we were not able to detect any impu- 
rity inthe water produced, we may consider the following num- 
bers expressing as exactly as possible the composition of this 
fluid. 
From the mean of three experiments it appears that 100 parts 
by weight of oxygen unite with 12.488 of hydrogen te produce 
water; which is equivalent to 88.9 percent. of oxygen, with 11.1 
of hydrogen: Whereas the number formerly assumed as the pro- 
portion of hydrogen to 100 of oxygen, is 13.27 instead of 12.488, 
which makes a difference of nearly a twelfth part. We can there- 
fore no longer doubt the reality of the error which we had sus- 
pected ; but it became necessary to examine the cause of it, by 
taking anew the densities of oxygen and hydrogen, which we 
performed in the usual methods, adopting, however, the follow- 
ing precautions, which appear to us of so much importance as to 
deserve a particular notice. 
_ It has been proved by Mr. Dalton, that no gas insoluble in wa- 
ter can remain confined in contact with this liquid, even for a 
short time, without absorbing a certain quantity of the gaseous 
mixture which water always holds in solution. When the density 
of the confined gas does not materially differ from that of atmo- 
spheric air, the addition of the gas which is absorbed from the 
water produces no material error; but where the confined gas is 
hydrogen, in particular, it is obvious that an alloy of no more 
than a hundredth part will produce a prodigious error in the 
estimated specific gravity. Itis very probable that to this cause 
(which was not known to Messrs. Biot and Arrago) we must 
attribute the error that affects the number whiclr they have given 
for the density of hydrogen. We have avoided it by covering 
the surface of the water that confines it with a stratum of fixed 
oil, which, as it is well known, makes the passage of the gas from 
the water much more difficult. We have operated and given the 
results of our experiments both on dried gas, and on gas satu- 
rated with moisture. One may employ either of them indiffer- 
ently, particularly when the external temperature is not very 
high. However, it has appeared to us that the observations 
made on the gases artificially dried accorded better with each 
other. Not that there is any uncertainty in the data, on which 
are founded the corrections that must be made for aqueous va- 
pour: 
