THE COMBINING VOLUMES OF HYDROGEN AND OXYGEN. 
413 
experiments in this series, a very large quantity of oxygen was liquefied and 
fractionated ; in others, only just enough for the purpose. The concordance was 
better than in series 1 and 2, the maximum deviation among ten consecutive 
determinations being less than 1 part in 14,000. 
The means of series 1, 2 and 3 differ by only 1 part in 100,000. The final mean, 
taking into account the probable error of each series, is 2'00292, with a probable error 
of less than ±4 parts in 1,000,000. The corresponding atomic weight for hydrogen 
is 1'00775. Since the oxygen was obtained from two totally distinct sources, and the 
hydrogen, though always derived from the electrolysis of barium hydroxide, was 
purified by two essentially different methods, the close agreement of the means of the 
three series suggests that the purity of the gases reached a high standard. Before 
accepting these values, however, it was necessary to consider the possibility of 
systematic errors inherent in the methods employed. One such constant error had 
already been detected. When a measured volume of gas was displaced from the 
pipette by mercury, a small bubble was always trapped in the dead-space, at the 
region of insertion of the glass point. Since, after the expulsion of the second 
hydrogen filling, oxygen was admitted without first washing out the pipette, the 
oxygen was contaminated with this small quantity of hydrogen. This implied 
a corresponding deficit of oxygen, the nett result being an excess of hydrogen 
in the residue after explosion equal to three times the volume of the original 
bubble.* 
By closing the pipette tap after removal of the gas, and lowering the reservoir, O, 
this bubble escaped into the vacuum so formed, and could be measured in the 
pipette capillary which was calibrated for the purpose. From a number of 
measurements with both hydrogen and oxygen the volume of the imprisoned bubble 
proved to be remarkably constant. As a mean of six experiments, agreeing to within 
1 c.mm.,this volume was found to be 6'5 c.mm., when corrected to normal temperature 
and pressure. This implied an excess of 19'5 c.mm., in the hydrogen residue, which 
would make the ratio of the combining volumes (and the atomic weight of hydrogen) 
too low by 33 parts in 1,000,000. The alteration involved was much smaller than 
the variations among the individual determinations, but the correction was justifiable 
since it was constant both in magnitude and sign. The corrected values for the final 
means become 2'00299 and 1'00778. 
Fourth Series. 
To test the validity of this correction we proceeded to carry out a further series of 
determinations in which the imprisoned hydrogen bubble was removed before the 
introduction of the oxygen. 
* The final hydrogen filling, which was contaminated in a similar way with a trace of oxygen, was 
removed by the mercury pump at the end of each determination. 
