620 PROCEEDINGS OF THE AMERICAN ACADEMY. 



above. As an overestimate of the loss of heat by lateral flow would 

 tend to an increase of the numerical value of v, the test here applied, 

 if it can be called a test, would seem to indicate that too great rather 

 than too small a correction had been applied for lateral flow in getting 

 the value —1019 X 10"^° for v. Reducing the applied correction by 

 one third of itself would reduce v at 159° from— 1019 X 10~^° to 

 something very near —948 X lO"^''. 



Further, and perhaps stronger, evidence tending to the same eff"ect, 

 that too large an allowance has been made for lateral flow in the work 

 of 1906 thus far described, is now at hand. During July, 1906, a re- 

 measurement of the temperature coefficient of thermal conductivity of 

 bars a and /3 was made^^ with calcined oxide of magnesium as a pack- 

 ing, through which very little, if any, convection of heat could occur. 

 The resulting value of this coefficient was— 0.00076, whereas the 

 value found when the asbestos fibre was used as a packing was 



— 0.00060. An overestimate of the allowance for lateral flow tends to 

 give too small a numerical value for this coefficient. Reducing the 

 applied correction, in the case of asbestos fibre, by one fifth of itself 

 would raise the value found for the coefficient from — 0.00060 to 



— 0.00074, which is very close to the value found with use of oxide of 

 magnesium as a packing. 



Circumstantial evidence, then, favors the opinion that the allowance 

 which we have made for loss by lateral flow through asbestos fibre 

 packing should be reduced about one fourth part of itself. Perhaps 

 the best we can do under the circumstances is to combine the various 

 estimates of the rate of change of v with rise of temperature as follows : 



Increase of v 

 per 1"^ Rise. 



From values of v at 32*' and 71° . . . -2.00 X lO"^"^ 



(1905) 

 From values of v at 137° and 182° . . -1.55 X 10""^ -2.0 X 10"" 



(1906) 

 From values of v at 52° and 159° . . -2.45 X lO-i*^] 



(1905 and 1906) 



Applying this rate of change to — 757 X 10"^", the value of v at 

 52°, as found in 1905, we get, as the value of v at 157°, — 971 X lO'^^. 

 This value is about 5 per cent less than — 1019 X 10~^*^, the value 

 at 159° as found from the measurements of 1906 taken alone. This 

 discrepancy is somewhat disappointing, and it must be admitted that 

 our values of v may be several per cent in error at any temperature. 



" By Mr. E. P. Churchill and myself. See the Appendix to this paper. — 

 E. II. H. 



Mean. 



