5 6 9 
Bonnier and Mangin for the Analysis of Air. 
Careful calibration of the tube showed that the error from slight 
inequalities of bore was between + o*oi and + 0*02 in the percentage of 
oxygen, and therefore practically negligible. 
The error did not appear to be due to incomplete absorption. In fact 
the volume of the residual nitrogen was observed frequently to increase 
after repeated contact with the absorbing film of pyrogallate of potash. 
This observation led me to investigate the effect of various purely physical 
changes, and particularly of the mechanical operations involved in the 
ordinary procedure. It was found that air introduced into the apparatus 
and brought repeatedly into the absorbing region of the tube soon began 
apparently to increase slightly in volume. This change was reversed, on the 
other hand, if the air was drawn back into the bulb. This empirical result 
is difficult to explain satisfactorily. There can be little doubt, however, 
that the apparent changes of volume are caused by movements of the film 
of moisture which wets the walls of the whole capillary. This film probably 
tends to accumulate in the horizontal graduated part of the tube, especially 
as this is at the highest level, a movement which would be accelerated by 
the encroaching films of absorbing solution. For some reason the transfer 
of air to the absorbing region tends to favour this accumulation, whereas 
withdrawal into the bulb tends to hinder it — indeed, to counteract it. 
Whether this be the true explanation or not, readings taken after with- 
drawal into the bulb are more constant than readings taken immediately 
after transfer to and from the absorbing region. 
Taking advantage of this observation and always reading the volume 
after withdrawing the air into the bulb, repeating the process a few times 
to make sure that the readings were constant, the following results were 
obtained, after correcting for temperature and introducing the slight calli- 
bration correction. 
Analyses of a Sample of Atmospheric Air. 
(I) 
( 2 ) 
( 3 ) 
(4) 
Average. 
%co 2 
0*04 
0*03 
0*07 
0*03 
0*04 
%0 2 
20*89 
20*89 
20*86 
20*88 
20*88 
% (CO a + 0 2 ) 
20*93 
20*92 
20*93 
20*92 
20*92 
These results agree with one another as closely as can be expected. 
The maximum difference from the mean percentages is 0-03, which corre- 
sponds to an error of little more than one-tenth of a small graduation less 
than a millimetre in length. 
The average percentage of oxygen, 20*88, is still somewhat lower than 
the 30*93 found by Hempel and by Haldane with more elaborate apparatus. 
In a recent paper Benedict 1 has given the results of analyses of atmospheric 
1 Loc. cit., p. 1 13. 
