PROFESSOR TAIT ON THERMAL AND ELECTRIC CONDUCTIVITY. 737 
BDB’=c, galvanometer=g. If, for instance, a=b, a=8, very exactly, and if 
we adjust B’ till there is no deflection, we have then — 
P=7; 
i.¢., AB and A’B’ have equal resistance. For accuracy by this method we must 
have, as THomson has pointed out, 7 very accurately, and ¢ very small. 
§ 19. The second and third methods which I employed require a differential 
galvanometer. This was very exactly adjusted, before the experiments, by 
putting the coils in multiple arc, and using the cell on them without a shunt. 
The exact balance was obtained by means of a box of resistance coils inserted 
in one or other of the branches. This being done, I connected one coil with 
A, B’, and the other with A’, B. Here the effect is approximately propor- 
tional to . 
(te oP") 
g 9g 

where g and g’ are the resistances in the galvanometer coils, and ¢ is the ratio 
of their deflecting forces on the needle when equal currents pass through them. 
The adjustment above described makes, very accurately, 
IJ =e, 
and the joint effect on the needle is therefore as 
a 
9 G2). 
Shifting B’ as before till there is no deflection, the resistances AB, A’B’ are 
equal, 
§ 20. But I find by trial, that by far the most expeditious and simple 
method is to connect the coils of the differential galvanometer directly with 
A, Band A’, B’.. Here the deflection is accurately proportional to 

25 a lvoe 
igre} 
so that the resistance ¢ is not involved. I found, in fact, that I could, without 
sensible alteration of the balance, put for c (which, in addition to short 
portions of the thick bars, was usually a brightly polished cube of copper 
of the same section as the bars, and clamped very tightly between them), a 
short thin wire, which became red-hot when the current was allowed to pass 
VOL. XXVIII. PART III. 9G 
