186 
DR. H. T. BARNES ON THE CAPACITY FOR HEAT OF WATER 
640 — 2 X 320 = a x — a 2 ; 320 — 2 X 160 = a . 2 — « 3 , &c., 
which should equal 0 if a l — a. 2 = <x 3 . 
If we let the error in 640 be equal to 0, then the error in 320 = jr (a l — a. 2 ) in 
terms of 640 even units, 
160 = \ — a. 2 ) — ct 2 — a 3 }, 
and so on for all the coils. 
The error in the bridge-wire, which we will call the bw. correction, is determined 
from the error in coil 10 obtained in terms of 640 even units. The calibration of the 
bridge-wire was done by inserting a small resistance, equal to about 3 centims. of 
bridge-wire, into the bridge circuit, so that by short-circuiting it by a heavy copper 
connector placed in mercury cups, the bridge-wire reading could be shifted the same 
amount at any part of the wire. The reading was found to vary '0005 centim. per 
centimetre on either side of the middle point, 19, in such a way as to increase 
towards 30 and decrease towards 0. This showed that the wire was slightly smaller 
towards the zero end, and lienee its resistance greater. As the equivalent length of 
10 centims., obtained in the calibration of the box coils, never occurred at exactly the 
same spot on the bridge-wire, there is a small correction to be applied to the values 
of « l3 «o, and « 3 , due to their position. The correction is worked out so as to reduce 
the values to a length of bridge-wire extending over the middle point, between 14 
and 24. The correction is very small, however, and would produce no appreciable 
error to the results if neglected altogether. In my own box the agreement of the 
equivalent length for the 10 coil above and below the middle point of the bridge-wire 
caused me to neglect this correction altogether. 
In Table XII. I give a complete series of readings taken to determine the coil 
corrections in the first box. In Table XIII. a summary of tests is given extending 
over a period of a year. 
Table XIV. contains the same obtained for my own box. The corrections in this 
latter case are somewhat larger. The reason being that it was more difficult to 
adjust the coils exactly when fused joins were used instead of solder, and at the 
same time preserve complete compensation. My aim was to be sure of having this 
latter condition fulfilled at the expense of the former, as the coil correction is always 
a definite and measurable quantity, and easy to apply. 
The signs are affixed to the corrections in the way they should be applied to the 
reading. The bridge-wire correction is given per centimeter of length. In taking 
the readings the galvanometer was used which has already been referred to in 
connection with the comparison of the standard resistances. The sensitiveness was 
obtained so as to give from 40 to 50 scale divisions per millimetre of bridge-wire on 
# 
reversing the current. For the small coils an external resistance of 350 ohms was 
