572 PROCEEDINGS OP THE AMERICAN ACADEMY. 



the average, perhaps as much as 1.5 per cent of the mean electromotive 

 force, and amounting in some cases to 2 per cent. This difference seemed 

 to grow greater in set (3) as time passed, for at the start, in July, it was 

 only about 0.5 per cent. This change may have been due to changes in 

 the method of experimenting or to some more or less damaging experi- 

 ences which the couples had suffered in some months of existence and 

 many days of use. It is difficult, especially when the german-silver 

 wire is not protected along its middle by tubing, to prevent an occa- 

 sional accidental pull which may stretch some part permanently a little 

 and thus change its thermo-electric quality. We know that a couple 

 made of hard-drawn german-silver with copper is 3 or 4 per cent more 

 powerful than a couple made of the annealed german-silver wire with 

 copper. 



On this point certain experiments made December 6 and 7, 1905, have 

 a bearing. Four pieces of carefully annealed german-silver wire were 

 each stretched to a permanent extension of 10 per cent in one part, and 

 then the stretched part of each was used with the unstretched part of the 

 same wire as a thermo-electric couple. The mean thermo-electric force 

 of the four couples thus formed was found to be about 0.6 per cent of 

 that of a couple made with copper and the annealed german-silver. 

 The stretching, which hardened the wire somewhat, thus bringing it 

 slightly toward the hard-drawn condition, changed its thermo-electric 

 quality slightly toward that of hard-drawn wire. If slightly different 

 annealing currents or slightly different periods of application of such cur- 

 rents can leave different pieces of wire as unlike as the stretched and the 

 unstretched pieces here tested, the observed differences of sensitiveness 

 of the couples A 3 , B 3 , C 3 , and D 3 may be due, in considerable measure, 

 to such inequalities. 



If we were now making a new set of couples, we should take more care 

 than we have usually taken heretofore to have all the pieces of german- 

 silver wire heated by the same strength of current and for the same 

 number of seconds. A current of 3 amperes applied for 5 seconds is 

 suitable for german-silver wire such as we have used. This wire is about 

 0.02 cm. in diameter, being No. 32 of the Brown and Sharpe gauge. 



Behavior of Manganin and of Constantin. 



The thermo-electric heterogeneity shown in german-silver by some of 

 the tests which have been described in the preceding pages induced us 

 to make somewhat similar tests with manganin and with constantin. 



A piece of annealed manganin wire 0.205 cm. in diameter was drawn 



