BETWEEN THE FREEZING AND BOILING-POINTS. 
179 
A thermometer placed in the hole extending through the middle of the embedded 
coil was taken as the temperature of the coil. As No. 4086 was arranged with 
a stirrer, a thermometer could he placed in the oil in contact with the coil, and the 
true temperature obtained. The current used on the bridge was not sufficient to 
cause j)erceptible heating. 
Table Till, contains the comparisons of the ohms, at an entirely different tempera¬ 
ture. These tests were made in the basement of the building, where the temperature 
was considerably lower than where the tests in Table All. were made. In this case 
also the coils remained at least a day or two at the temperature of test, and did not 
vary to any extent from that. The table is arranged as in Table VII., only the com¬ 
parisons were made in terms of No. 3569. This shows a very good agreement of all 
the platinum-silver standard ohms, including the Reichsanstalt ohm, 1214, but shows 
that by assuming the corrections of 3569, the value of 4086 is very much above that 
given in its certificate. This difference indicates an error of ‘00059 ohm assuming 
3569 as correct, or ‘00065 referred to the mean of all the ohms. In Table VII. 
we saw that the platinum-silver ohms were all lower than their certified values 
when calculated assuming 4086 to be correct, the mean difference being ‘00052. 
These differences are both in the same direction as regards the relationship of 4086 
to the other ohms. The difference of 1 in 10,000, obtained by Mr. Fraser and 
myself between the two values, i.e ., ‘00065 and '00052, must be ascribed to the wide 
difference in temperature of our respective tests, as well as to the uncertainty ot 
knowing accurately the true temperature of the paraffin-embedded coils. 
We are forced now either to accept the certificate of 4086, and reject all the other 
11 ohms as being in error, including the Reichsanstalt Standard, or to reject the 
certificate of 4086, and accept the certificates of all the others. The alternative of 
giving 4086 equal weight in the mean seems to be hardly justifiable considering the 
mass of evidence against it. 
I have decided to reject the certificate of 4086, and I have accordingly corrected 
it in the following way : in terms of the platinum-silver standard ohms, 4086 is 
equal to its certified value +‘00052 by the comparison made at 22° C. By the 
comparison made at 13° C. it becomes equal to its certificate +‘00065. By 
comparing directly with 1214, the value of 4086 becomes equal to its certificate 
+ ‘00056 in one test, and +‘00061 in another test, or equal to +‘000585 in the 
mean. This agrees very closely with the mean value of the two separate determina¬ 
tions with the other ohms, which comes out +'000585. We may, I think, then 
safely assume that the value of 4086 is equal to its certified value +‘00058, which 
comes out ‘99978 + ‘00058 at 15°‘9, or 1‘00036 + ‘000018 (20° - 15°‘9), or equal to 
1 ‘00043 true ohms at 20° C. 
A summary of the various comparisons made of the two new platinum-silver ohms 
is given in Table IX. in terms of 4086, assuming for convenience that it is exactly 
1 ohm at 20° C. The resistance of each ohm is reduced to 20° C. in column 4 of each 
o A o 
u A u 
