320 
Proceedings of Royal Society of Edinburgh. [sess. 
On the Gravimetric Composition of Water. A Preliminary 
Communication. By W. Dittmar. 
(Read February 3, 1890.) 
On the strength of Dumas’ famous Recherche sur la Composition 
de VEau * and adopting the great master’s own interpretation of his 
results, all chemists, until lately, agreed in assigning to the atomic 
weight of oxygen the value 0 = 16 (H=l); and it is on the 
strength chiefly of the same experiments that many of us now hold 
that 0 = 15*96 is a closer approximation to the truth ! 
In these circumstances it surely is worth while to look into 
Dumas’ work with the help of critical experiments, and try to see 
whether he was not right in thinking that — all his great efforts 
notwithstanding — the difference lies within the influence of his 
method-errors ; the more so, as all these errors (as far as one can see 
without experimenting) tend to raise the experimentally ascertained 
value of the ratio H : 0 above its true value. 
I accordingly, some time ago, caused my private assistant, Mr 
Henderson, to join me in this inquiry, and, thanks to his youthful 
energy and indefatigability, we have already made considerable pro- 
gress in our work, and hope before long to lay a complete account 
of it before the Society. Meanwhile, I content myself with stating 
that we have succeeded in so modifying Dumas’ modus operandi as 
to give a higher degree of constancy to the weighings, and to reduce 
the trouble and loss of time involved to far less than it was with 
the original form of the method. 
The principal object of the present notice, however, is to direct 
attention to an oversight which Dumas made himself guilty of, and 
which, as far as I am aware, has never been noticed before. What 
I allude to is that Dumas, while weighing his oxygen (virtually) in 
vacuo , weighs his water in air, and forgets to reduce this latter weight 
to the vacuum. 
That the correction tells very considerably upon the calculated 
weight of the hydrogen a very little reflection is sufficient to show ; 
* Ann. Chim. Pliys. (3), vol. viii. p. 189. 
