for some Pure Metals. 221 



enormously high (as thermoelectric forces go), as in the case 

 of half a dozen different metals I observed *3 volt, about the 

 same as that given by a junction of bismuth and antimony at 

 that temperature, did the ordinary formulae hold so far. At 

 lower temperatures these irregular effects are not so great, 

 being more than proportionately smaller, but still on several 

 occasions I found them distinctly marked even as low as 

 200° C. 



In any series of observations the relation between the 

 observed values of the temperature T and E.M.F. e is of the 

 form e=a(T— T ) +6(T — T ) 2 , where T is the temperature of 

 the cold junction and a and b are constants to be determined, 

 b being positive or negative, according as the lines of the two 

 metals being examined intersect below or above T . In 

 each, a and b were determined by the method of least squares. 

 e having been calculated again in each case from the values 

 of a and b thus obtained and the observed values of T, the 

 means of the differences between the values of e thus 

 calculated and those observed are given below in each case as 

 the mean error in absolute units. To compare the error in 

 different cases, I have also expressed it as the mean error in 

 temperature. It will be noticed that frequently the dis- 

 crepancies between different sets of observations on one junc- 

 tion are enormously greater than the mean error in one set, 

 in spite of the fact that, as far as one could tell, they were 

 in exactly the same condition in each case and were not even 

 touched between the experiments. In view of the experi- 

 ments referred to above and these discrepancies, one is led to 

 the suspicion that the thermoelectric constants are not really 

 constant, but that they vary in a given specimen in a manner 

 which, if not arbitrary, yet arises from changes in condition 

 which are inappreciable. I usually made three sets of obser- 

 vations on any pair of metals I examined, but with copper and 

 lead I made more, — first to find out if the constants tended to 

 become more constant, or, if not, to see if they varied in any 

 regular manner ; and, secondly, because I had taken copper as 

 a base-line for some metals which lay close to it, and it was 

 therefore necessary to determine its position with more than 

 usual accuracy. The following are the thermoelectric heights 

 observed ; the metal with the higher line being in each case 

 put first : — 



