86 On Ascertaining the 
from D to E is occupied by mercury, and the remainder of 
the bulb E F'isa Torricellian vacuum, the tube being hermeti- 
cally sealed at both ends to exclude barometric influence. 
The effect of heat will be to expand the air in the bulb B, 
and by the increase of its pressure to force the mercury out 
of the bulb C D into EF, the upper part of which being 
vacuum will offer no resistance. The weight being thus 
removed nearer to the centre of oscillation, will be tanta- 
mount to a shortening of the pendulum, and will cause it to 
vibrate more rapidly, and in exact proportion to the tem- 
perature, as it is well known that the expansion of air is uni- 
form with every increment of temperature. 
A drop of oil on the surface of the mercury in C D, and 
the substitution of hydrogen gas for air in B would probably 
add to the perfection of the instrument ;—the object of the 
oil being to prevent the transfer of air into the vacuum, and 
the hydrogen to obviate the action of common air on the oil. 
By a slight and obvious modification of this instrument a 
sensitive air thermometer may be constructed. 
It is, however, obvious that the true mean temperature is 
not necessarily the same.as the mean of the two extremes of 
the twenty-four hours. The thermometer may have been for 
many hours near its highest point, and for a short time only 
near its lowest, or vice versd. The force of this objection 
becomes considerable in a climate like this, where a change of 
temperature to the extent of thirty degrees is not unfrequently 
known to occur during a single hour. Admitting, therefore, 
the value of Six’s thermometer, as registering in a most con- 
venient manner the extremes of heat and cold, it cannot, I 
think, with propriety be depended upon, as an instrument 
affording data from which to calculate mean temperatures, 
except approximately; and this approximation may be 
seriously remote from the exactness which modern science 
demands. 
I now proceed to describe the principle of the instrument 
which I propose for ascertaining mean temperatures. It is 
well known that the pendulum of a clock vibrates more or 
less rapidly according to its length; that the pendulum is 
elongated by heat and shortened by cold; and consequently, 
that an ordinary clock has a tendency to go slower or lose 
time in warm weather, and to gain time in cold. The effect 
of slight changes of temperature upon an ordinary pendulum 
is very inconsiderable; but, if we can succeed in constructing 
a pendulum, which shall be highly sensitive of heat and cold, 
