HALL. — THERMAL AND ELECTRICAL EFFECTS IN SOFT IRON. 619 



It must be confessed that this assurance is a little difficult to hold. 

 The close scrutiny which we have given to the matter of lateral flow 

 has left a rather troublesome doubt as to the soundness of our working 

 assumption that this flow is not to any important extent a function of 

 the longitudinal gradient of temperature. Upon this point we must 

 bring to bear whatever further evidence is available. 



A satisfactory mode of investigation, if it were not so extremely 

 laborious, would be to repeat the determination of v wnth other kinds 

 or states of packing than those here used. Something like this varia- 

 tion of conditions we find in a comparison of our 1905 work with 

 that of 1906, or in the former cotton- wool, closely packed in, was 

 used instead of the rather loosely packed asbestos of 1906, and it has 

 already been stated that the coefficient of lateral outflow, at 100", 

 per degree difference of temperature between the main bars and the 

 guard-ring bars, was about 1.7 times as great with the asbestos as 

 with the cotton. If we have made any large error in using the as- 

 sumption now on trial, we should expect a marked discordance be- 

 tween the results obtained from the older and the more recent work. 

 But as the two do not cover the same temperature, one lying between 

 13° and 90°, while the other lies between 115° and 204°, how can we, 

 without begging the main question which this research deals with, tell 

 what is accordance and what is discordance 1 



In this difficulty we get some help from the fact that in each case 

 we have taken note of the diff"erence of gradient at the middle cross- 

 section, c, of the main bars, so that it is possible to make from the 

 data of each year's research an estimate of the rate of change of v with 

 change of temperature, which estimate should not be greatly affected 

 by error in regard to the absolute magnitude of the loss by lateral 

 flow. These estimates are : 



From Data of Range of Temperature. '^"ilUe'of Temp^^^ ^®'''°* 



1905 32° to 71° -2.0 X 10"^° i_i q x 10-1° 



1906 137° to 182° -1.6 X 10"" J 



If now, taking —1.8 X 10"^° as the rate of change of v from the mean 

 temperature, about 53°, of the 1905 work to the mean temperature, 

 about 159°, of the 1906 work, we reckon from the value of v at 53° 

 what its value at 159° should be, we get 



-757 X 10-^^ -(159-53) 1.8 X 10"^° = -948 X 10-^°. 



The value found from the data of 1906 is, as we have seen at page 610, 

 — 1019 X 10"^", about 7 per cent larger than the value as estimated 



