64 
MESSRS. G. F. C. SEARLE AXD T. G. BEDFORD 
corresponding to may ditier somewhat from the value Bj found by the 
galvanometer. 
The experiments of §§ 45-49 below show that, for a given value of Hq, W may vary 
considerably from causes depending upon the magnetic history of the iron even when 
its temperature and strain remain constant. Thus, unless it is possible to give an 
accurate account of the history of the iron and of the manner in which the magnetic 
force changes from Hq to — Hq, it is im})ossible to assign any definite meaning to W, 
and hence, in the absence of such an account, it is useless to attempt a very close 
comparison between the value of U found hy the dynamometer and the value of W 
found from the cyclic B—H curve. But the agreement between the two values as 
recorded in the table will perhaps suffice to give general confidence in the method. 
§ 41. Test hy Variation of the Resistance of the Secondary Circuit. —The process of 
comparing the value of U found by the dynamometer with the value of W found 
from the area of the cyclic B—H curve is laborious, and after all does not furnish an 
absolute test of the equality of U and W for that particular law of variation of the 
magnetic foi'ce which is obeyed when the measurement is made, as was explained in 
§ 40. Wlien, as was the case in most of our experiments, the specimen is a fine wire 
not exceeding 2 millims. in diameter there is a second method, described in § 14—it 
may be termed, in contrast with the first, a self-contained method—which is easy of 
application, enalding us to test the accuracy of the dynamometer measurements by 
the dynamometer itself If, in any case, we can find the effect of Y, have a 
superior limit to the effect of X, for lyy (14) X/Y < QS^i'cr. We now give the 
results of some applications of this second method. 
It will generally suffice to take two values of S, one doidjle the other. Since, when 
S is varied, <h is inversely proportional to S, it follows that U [= NOT (dj+ O.f Aiaf)] 
is proportional to S ff- O.f provided that (f) and 6 are corrected for damping (§ 14). 
Hence, in testing the accuracy of the dynamometer measurements, it is sufficient to 
compare the values of S {6^ + do) for the two values of S. If S be doidjled the 
consequent diminution in S {0^ -ff ffi) is, by § 14, equal to the amount wliich must be 
subtracted from the product for the larger resistance in (_)rder to obtain the value of 
the product corresponding to S = co . 
(a.) We will first refer to the results obtained by one of us in November, 1895, for 
an iron wire '0201 sq. centim. in section, with Hg = 8’34 and Bg = 10170, the value of 
U being 17780 when S = 59‘7 ohms, while W was found from a cyclic B—H curve to 
be 7409. A sinq)le mercury rocking key was used, and the primary circuit had only 
the self-induction of the dynamometer and the solenoid. As S rose from 59’7 to 
584 ohms, U fell from 17780 to 14550, being closely represented by U= 14070-1- 
221900/S. From this we find by (20) that when S = 59‘7 ohms, Y — 3720. In these 
experiments n = 1285, I = 11‘5 centims., G = '785 sq. centim., and hence if L and 
M (§ 11) be neglected, X/Y = DOG vdien S = 59‘7 ohms. Thus Z = X -ff Y = 
(D66 -1- 1)3720 = 9890, Avhen S is 597 ohms. Subtracting this from the corre- 
