1875.] The Mechanical Action of Light. 339 
with the pump, so as to readily admit of being sealed off 
at any stage of the exhaustion. pb is the pump-gauge, 
and £ is the barometer. 
The whole being fitted up as here shown, and the appa- 
ratus being full of air to begin with, I passed a spirit-flame 
across the lower part of the tube at 0b, observing the 
movement by a low-power micrometer: the pith-ball (a) 
descended slightly, and then immediately rose to consider- 
ably above its original position. It seemed as if the true 
action of the heat was one of attraction, instantly overcome 
by ascending currents of air. A hot metal or glass rod and 
a tube of hot water applied beneath the pith ball at 0 pro- 
duced the same effect as the flame ; when applied above at a 
they produced a slight rising of the bail. The same effects 
take place when the hot body is applied to the other end of 
the balanced beam. In these cases air-currents are suffi- 
cient to explain the rising of the ball under the influence 
of heat. 
In order to apply the heat in a more regular manner, a 
thermometer was inserted in a glass tube, having at its 
extremity a glass bulb about 1} inches diameter; it was 
filled with water and then sealed up (see Fig. 2). This was 
Fic. 2. 
arranged on a revolving stand, so that by means of a cord I 
could bring it to the desired position without moving the eye 
from the micrometer. The water was kept heated to 70°C., 
the temperature of the laboratory being about 15° C. 
The barometer being at 767 millims. and the gauge at 
zero, the hot bulb was placed beneath the pith. ball at 0b. 
The ball rose rapidly. The source of heat was then re- 
moved, and as soon as equilibrium was restored I placed the 
hot-water bulb above the pith ball at a, when it rose again, 
—more slowly, however, than when the heat was applied 
beneath it. 
The pump was then set to work; and when the gauge 
was 147 millims. below the barometer, the experiment was 
tried again: asimilar result, only more feeble, was obtained. 
The exhaustion was continued, stopping the pump from 
