612 PROCEEDINGS OF THE AMERICAN ACADEMY. 



stands for the temperature coefficient of this property between ti and 

 tn, we have 



^•ft ^ (Jh = h X gi 



and / = ^A^' ^^t^-t,)=(^f^-l^-r- {t, - O 



= [^ -\\^{t„-t^= -0.00060+. 

 \9h J 



As the value of this temperature coefficient is a matter of consider- 

 able interest, and as such coefficients are notoriously difficult to de- 

 termine, it seems worth while to inquire how much the two values 

 differ which are obtainable from the data of February 7 and the 

 data of March 2 and 3, taken separately and corrected, as well as pos- 

 sible, for the small displacements of the various junctions from their 

 nominal planes of cross-section. 



Measurements, which were capable of no great accuracy, have indicated 

 that the junctions on the guard-ring bars are placed, on the average, 

 about 0.15 cm. farther toward the east end of the apparatus than the 

 con*esponding junctions on the main bars. This displacement, with 

 the gradient of temperature ordinarily prevailing in the bars, about 5.5, 

 would make a difference of about 0.15 X 5.5, or 0.83, degree. The values 

 of ^ — t', corrected accordingly, would be 2.20 + 0.83 = 3.03 for Feb- 

 ruary 7, and 3.27 — 0.83 = 2.44 for March 2 a,nd 3. The values of/ 

 worked out from the two sets of data, as thus amended would be very 

 nearly —0.00047 and —0.00073 respectively. 



The value of / found in our observations of last year on the same 

 bars of iron, between 13° and 87°, was —0.0007, the value of k at 13° 

 being the basis. The values of/ found in this paper are for the interval 

 from 115° to 204°, approximately, and the value of k at 115° is here 

 taken as the basis. If we use the value oik at 13° as our basis now, 

 we get approximately —0.00056, instead of —0.00060, as the value of/ 

 between 115° and 204°. The work on thermal conductivity described 

 in the present paper is probably entitled to more confidence than that 

 described in our paper of last year, though the correction for lateral 

 flow was very much greater this year than last year. 



On the whole, in view of the acknowledged difficulty of making any- 

 thing like an accurate determination of/ the general accord of all our 

 results in this phase of our work is reassuring.^ 



' See the Appendix to this paper for an account of further experiments on the 

 temperature coefficient of thermal conductivity in bars a and /3. 



Jaeger and Disselliorst found (see Beib. zu den Ann. der Tliys. for 1901, p. 20) 



