52 Professor CUMMING on the Developement 
soldered together, with a thin plate of copper interposed, they no 
longer act as one, but as two distinct bars. When heat is applied 
at the extremity a, the deviation is, as usual, negative; at e¢ the 
same; but if at 6, the deviation is positive, the extremity a be- 
coming the cooler end, and the part cd merely conducting the 
electricity: but as the bar cools, ab, and ¢ the extremity of the 
other part, gradually assume the same temperature, and conse- 
quently the bar acts negatively as at first. It appears then, that 
when the bar was entire, the heat was not merely conducted from 
one extremity to the other, but, by some means modified in its 
progress, and that, for the production of this species of electricity, 
there is required the juxta-position of two particles of the same 
metal at different temperatures. If therefore, a cylindrical bar, 
unequally heated, be supposed to be divided into an indefinite 
number of circular laminz, each will act, as a layer of hot particles 
upon the lamina on one side, and of cold upon that on the other, 
and the total effect of the bar will arise from the aggregate action 
of these lamine. 
By soldering wires to a long rod of bismuth (Fig. 7.), the parts 
of which were alternately hot and cold, it was found that the 
action of the whole exceeded that of any two portions taken se- 
parately, and as the only condition appears to be, that there should 
exist a difference of temperature between two adjoining particles, 
it may be inferred, that if it were possible to increase these divisions 
sine limite, each bar would act as an assemblage of an indefinite 
number of small plates; as the common magnet may be conceived 
to be composed (if the expression may be allowed) of an indefinite 
number of atomic magnets.* 
* The connection of the wires may be either 1 and 2, 2 and 3, 3 and 4, or 1 and 4, 
or 1 and 3 may be placed in one cup of the galvanoscope, and 2 and 4 in the other, in 
which case the effect is the greatest. The deviation caused by connecting 1 and 4 is not 
affected by connecting at the same time 2 and 3; this seems unfavourable to the sup- 
position of circulating currents. 
