519 
and magnetic rotations. 
For the purpose of determining whether the air, heated by 
the upper surface of the metal disc, acted on the needles, as 
it does on the vane of a smoke-jack, I made the following 
experiment. 
Experiment 21. 
Having formed a needle of sheet brass, thinner and lighter 
than those hitherto employed, but of the same form, I bent 
the circular parts, at the extremities, in such a manner, that 
they formed angles with the longitudinal axis of the needle ; 
one of about 5 0 , the other about 15 0 . This being balanced 
above the copper plate, the lamp was lighted below it, but 
no rotation was communicated to the apparatus. The needle 
soon began to acquire vibrations in a vertical plane ; but 
after five minutes' exposure there was no perceptible rota- 
tion. I now bent the circular extremities of the needle so 
as to form angles of about 35 0 and 40°. On replacing it, and 
lighting the lamp, the needle moved about 23 0 , in a contrary 
direction to that which the impinging air ought to have pro- 
duced ; it then returned, and went on through several revo- 
lutions in the opposite direction. The needle employed in 
this experiment projected a little beyond the copper plate, and 
would therefore receive some of the heated air passing from 
the lamp along the under side of the plate, to which cause 
the vertical vibrations may perhaps be ascribed. 
From this experiment it appears, that if the needles are 
made flat with ordinary care, rotation will not ensue from 
the action of the heated air. 
I shall now briefly restate the various causes to which 
these rotations might be ascribed, and refer back to the 
experiments which refute each supposition. 
