CONTINUOUS ELECTRIC CALORIMETRY. 
91 
The box was originally intended for working chiefly with thermometers of pure 
platinum having a zero resistance equal to the largest coil, i. e ., 2560 box-units, and a 
fundamental interval of 1000 box-units, or nearly 10 ohms. This gave a very 
convenient scale of 10 centims. of the bridge-wire to the degree, and made it possible 
to take readings to the ten-thousandth part of a degree. Thermometers of double 
the resistance, having a scale of 20 centims. to the degree, were employed for some 
differential work, e.g., for demonstrating the lowering of the melting point of ice due 
to one atmosphere of pressure ('0075° C.) at the May Soiree of the Royal Society in 
1893, where this box was first exhibited. But for ordinary temperature measure¬ 
ments, unless the insulation were extremely perfect, it was found that very little 
could be gained in point of accuracy by going beyond 10 ohms for the fundamental 
interval. For the boiling-point of sulphur, a fundamental interval of 5 ohms was 
found to be preferable ; and 1 or 2 ohms for higher temperatures where the insulation 
was necessarily less perfect and the conditions less steady. As the portion of the 
bridge-wire actually utilised never exceeded 5 centims., and averaged only 2’5 centims., 
or \ of a degree, it was seldom necessary to take any account of the calibration 
correction of the bridge-wire, the errors of which proved to be less than one part in 
500, or one two-thousandth part of a degree, without correction. In the best 
mercury thermometers it is unusual to calibrate closer than 2° intervals, and the 
corrections are necessarily uncertain to two or three thousandths, even if the inter¬ 
polation formulae can be trusted to one part in a thousand, which is very doubtful. It 
is easy to see how great is the advantage of the platinum-thermometer in point of 
ease and accuracy of calibration. 
The comparison of the 9 coils of the resistance box, which were arranged on the 
binary scale, could be carried out and the relative values calculated in less than an 
hour. The peculiar advantages of the binary scale for this purpose are frequently 
misunderstood or misrepresented. The most important of these advantages are : 
(1) That a given resistance can be represented only by one particular combination of 
coils, so that there can never be any doubt as to which combination was employed 
for any given reading ; (2) That the least possible number of separate coils are 
required, and that the complete calibration requires the least possible number ot 
readings, and can be effected in the least possible time. It is often urged as an 
advantage of less simple and symmetrical arrangements, that each resistance can be 
made up in a great variety of ways, which act as checks in case of doubt; and that 
the relative values of the coils can be compared in a number of different combinations, 
so that several equations can be obtained for evaluating each resistance. This is, no 
doubt, an advantage in cases'where the accuracy of calibration depends on micrometric 
estimation, as in calibrating a mercury-thermometer. But, if a good galvanometer is 
available, the accuracy of comparison of resistances is not limited in this way. We 
may, therefore, fairly consider the disadvantage of the excessive expenditure of time 
in the observations and calculations ; moreover, the risk of error due to changes of 
N 2 
