BETWEEN THE FREEZING AND BOILING-POINTS. 
185 
wire was 40 centims., but it could be read only between 6 and 34 centims., leaving a 
margin of 6 centims. at each end. 
Professor Callendar was kind enough to allow me to make a resistance box after 
this design. This box I have used in my later determinations of the specific heat. 
It differs from Professor Callendar’s box in having a slightly greater unit, i.e., 
1 centim. of bridge-wire equal to ‘0095 ohm, and the coils were made from bare wire 
wound on mica frames and annealed. Solder joins were avoided between the wire 
forming the resistance coils and the copper connecting wire, by fusing directly to the 
copper. 
Each of the larger coils, before putting in place in the box, was tested for 
compensation in a specially constructed oil-bath, the temperature of which could be 
changed quickly at will in a way similar to the paraffin-bath used in the standard 
resistance determinations. Each coil was also made of either two or three wires in 
parallel, ‘15 mihim. in diameter, so as to avoid current heating. They were specially 
designed for immersion in oil when in place in the box, but this was not found necessary. 
It was not deemed necessary to test the small coils, from 10 to 40, for compensation, 
as the test of the larger coils showed that the calculation of the lengths of wire 
necessary was so nearly correct as to leave little room for error in the smaller coils 
over a wide range of temperature. The ratio coils in the box were made from 
'15-millim. platinum-silver wire wound in parallel on a mica frame, and were adjusted 
to equality on the Thomson-Yarley slide box. The resistance coils were connected to 
mercury cups and short-circuited when not in use by thick copper connectors. 
The calibration of these boxes consists in determining the errors in the different 
box coils and the calibration of the bridge-wire and scale. 
In determining the total change in resistance of the thermometers between 0° and 
100°, which is termed the fundamental interval, or briefly F.I., the largest that it 
was necessary to use was coil 640. It is evident that, provided this coil is accurately 
compensated, it is the best one to which to refer the F.I. It is entirely unnecessary 
to know its absolute value in ohms provided we assume it equal to 640 even units, 
and refer the other coils, including the bridge-wire, to it. 
From 640 down every coil differs from the sum of all the rest by very nearly 
10 centims. of bridge-wire, or the size of the smallest coil. If we compare the lengths 
of bridge-wire obtained by differencing the coils in this way, we obtain the usual 
series of equations of the form 
640 — sumq = a Y ; 320 — sunn = a % ; 160 — sum, = n 3 , &c., 
where oq, c< 2 , and a 3 are very nearly 10 centims. and involve the coil errors. 
If we eliminate the sum from any two equations, remembering that the next 
lowest sum differs from the one before by the lesser coil, then we have a series of the 
form 
2 B 
VOL. CXCIX.—A. 
