604 
Proceedings of the Royal Society 
For any deflection 0 , if the length of the small magnet he negli¬ 
gible compared with x, we have 
E sin. 0 = m 
cos. 6 
.*. in a tan. 6 . 
[This simple formula holds, of course, however complex he the 
distribution of magnetism in the large magnet, provided the rela¬ 
tive intensities of magnetization at different parts, and their direc¬ 
tions, remain unchanged by heating.] 
Disturbances were experienced in the form of thermo-electric 
currents in the pot and brass ring supporting it (these acted against 
one another), hut their effects were rendered insignificant by remov¬ 
ing the flame, and allowing the whole to come to a uniform tern- 
perature before reading. The direction of these currents, and there¬ 
fore that of the disturbance to which they gave rise, could he re¬ 
versed by changing the position of the flame relatively to the pot; 
hut a smaller disturbance of a more unaccountable nature presented 
itself during the heating of the pot, which did not depend on the 
position of the flame, and could not be got rid of. This latter 
disturbance, which increased with the temperature, resulted in a 
gradual alteration of zero, and in consequence the deflections, cor¬ 
responding at least to the higher temperatures in the curves and 
all the ordinates of the lower part of curve III., are somewhat less 
than they ought strictly to be. 
Curves I., II., and the upper part of curve III., show how the 
absolute magnetism diminishes as the temperature of the magnet 
increases ; the lower part of curve III. shows how the magnet re¬ 
gains its power when the temperature again falls, and it is seen at 
once from it that, when the magnet is allowed to cool after being 
heated, the deflection corresponding to a given temperature is less 
than that obtained at the same temperature when the magnet is 
being heated, thus indicating a loss of magnetic power, and the 
difference of the two deflections is greater the lower the tempera¬ 
ture. It is principally on this account also that the curves I. and 
II. do not coincide, for the experiments were performed on succes¬ 
sive da} T s, and it was found that that magnet took about two days 
after such heating to acquire its original power. The magnet used 
