88 PROCEEDINGS OF THE AMERICAN ACADEMY. 



But it is unlikely that R for the electron is as great as R g for a gas 

 molecule at ordinary temperatures. 



If we take R = 0.5R g = 68.5 X 10" 18 , we get 



from (22), v = 0.54, and from (22'), v= 1.54. 



If we take R = 0.1R g = 13.7 X 10" 18 , we get 



from (22), v = - 1.32, and from (22'), v = -0.32. 



If we put v = 0, we get 



from (22), R = 32 X KH 8 , and from (22'), R = 16 X 10~ 18 . 



As negative * -values of v seem improbable, the indication of this 

 investigation is that in cobalt, at the temperature for which a = 2000, 

 R may well be as small as 0.25 R g , but is probably greater than 0.1 R v . 



Turning now to antimony we have 



a = -(\ + | - vjR = ~ 1000 (23) 



or <r = -(1 + p- v)R= - 1000 (23') 



If we take R = R g , we get 



from (23), v = 1.12, and from (23'), v = 2.12. 

 If R = 0.5R g , we have 



from (23), v = 1.23, and from (23'), v = 2.23. 

 If # = 0.1iJ ff , we have 



from (23), v = 2.16, and from (23'), v = 3.16. 



Here also, since large values of v seem rather improbable, the indi- 

 cation appears to be that R is greater than 0.1 R g . 



On the whole, thus far, it appears that the hypothesis of combined 

 action of (A) electrons and (B) electrons, with the former playing 

 much the greater part in electric conduction and th® latter having a 

 very important function in a metal not uniformly heated, gives a 

 reasonable account of the phenomena of electric conduction, 21 includ- 

 ing change of conductivity with change of temperature, and of the 

 Thomson effect, in pure metals. 



It is now time to consider how well the same hypothesis will apply 

 to the phenomena at the junction of two metals, the Seebeck effect 

 and the Peltier effect. 



21 It seems not unlikely that the change of resistance of metals in melting, 

 in most cases an increase, is due mainly to the change of volume, which also 

 in most cases is an increase. See Appendix for data and discussion. 



