264 
MESSRS. R, T. GLAZEBROOK AND E. B. SARGANT ON THE 
The difference, amounting as it does to about 14 parts in 100,000, is obviously less 
than the probable error of our result, and there can be little doubt but that the value 
•9866 is only two or three parts in 10,000 from the truth. This value, as Lord 
Rayleigh has pointed out, is strikingly confirmed by Joule’s latest determinations 
of the mechanical equivalent of heat. 
In conclusion, we would discuss further one or two possible sources of error. We 
have assumed that the ratio of Y/S is independent of the temperature. This is not 
true, for S consists of two parts; one part of about 3000 units has the same tem¬ 
perature coefficient and is always at the same temperature as Y, so that for it our 
assumption is justified ; the other part of the galvanometer, of about 60 units, is 
of copper, of which the temperature coefficient is ’003, and it is not at the same 
temperature as V. 
Now the whole range in the temperature of the galvanometer is about 3°, the range 
on either side of the mean 1 0, 5, and an error of 1°’5 in the temperature of the galvano¬ 
meter we can show would affect R to the amount of about one part in 8000. Thus 
those values of R for which the temperature of the galvanometer differs most from 
the mean may be affected to the amount of 1 in 8000 by this source of error. Taking 
the mean of all our observations, however, the error vanishes. 
Professor Chrystal had warned us that he had had some difficulty in securing 
sufficiently good insulation between the wire of the coils A and B and the brass rings 
in which they were wound. We therefore tested both coils to see that this was 
maintained. One pole of a battery of 25 Leclanche cells was connected with the 
brass ring, while the other pole was put in contact with one electrode of a galvano¬ 
meter of 2000 ohms resistance, the second electrode of the galvanometer being 
connected with the wire of the coil. 
Deflections of 60 and 80 divisions of the scale were observed for the two coils A and 
B respectively. The same battery through 100,000 ohms when the galvanometer is 
shunted with 8 ohms gave a deflection of over 100 divisions, the insulation resistance 
therefore is considerably over 
100,000 x2000 
8 
or 
25,000,000 ohms. 
The error that might arise from the use of a paper scale has been discussed. In the 
calculations corrections have been applied to the scale readings to reduce them to the 
standard metre. The corrections taken were the mean of four series of observations— 
two by R. T. G., two by E. B. S.; and these, though made at very different times, 
varied only by quantities comparable with the error of an observation. 
The most serious objection, however, that can be raised applies to all observations 
in which a ballistic galvanometer is used, and there were two points here which 
seemed to require special notice. Is it right to assume that on reversing the primary 
