430 On Jolly’s Hypothesis as to the Cause of the Variations 
trustworthy we must abandon the hypothesis, or else attribute 
to its supposed cause a magnitude altogether incredible. I will 
examine some of the experimental evidence, that the oxygen 
in the atmosphere sometimes falls below the mean by as much 
as 0°004 or 0:005. 
I will not cite any analyses made before the year 1841. In 
that year Dumas and Boussingault found it necessary to resolve 
by experiment the doubt as to whether the true proportion of 
oxygen in the air were exactly one-fifth, or were about twenty- 
one per cent, or were a variable quantity. If this was the un- 
certainty as to the mean cf multitudes of analyses, it is obvious 
that we can by no means attribute to a single analysis a degree 
of precision sufficient to aid in the present inquiry. But in 
that year, Dumas and Boussingault used a new method of 
analysis, by means of which sufficient accuracy was obtained, 
and proposed an elaborate system of analyses on air collected 
simultaneously at different places. Lewy went to Copenhagen 
to take part in this system, carrying with him apparatus from 
the laboratory of Dumas and Boussingault. He had the codp- 
eration of Oersted, and his results were communicated to the 
Academie des Sciences by Dumas and Boussingault. Four of 
his results on four samples of air, collected at sea on the voyage 
” Copenhagen, showed a proportion of oxygen as low as 
‘20-45. 
Regnault’s results will command entire confidence. A sam- 
ple collected in the Bay of Algiers, June 5, 1851, gave 0°2042 
and 0°2040 oxygen. A sample collected in the Bay of Bengal, 
1879, gave 0°2045. The other analysis of this sample was lost 
by the accidental use of hydrogen containing a little air. But 
even this analysis, which of course gave the proportion of oxy- 
gen too high, gave only 02049. : oe 
The analyses of Macagno, at Palermo, made by absorbing 
oxygen with pyrogallol, I forbear to cite. 
