362 PROCEEDINGS OF THE AMERICAN ACADEMY. 



SO that a correction had to he made for the heat thus carried off. A 

 knowledge of the temperature gradients at the ends of the section and the 

 thermal conductivity was required for the application of this correction. 

 Different observers have obtained very different values for the thermal 

 conductivity of copper ; therefore, in order to see what effect an error in 

 the estimation of this quantity would have on the final value for the 

 Thomson Itlffect, the part of the work depending on this has been worked 

 through with two different values. 



Precautions against Thermo-electric Effects at the Junctions of Dissimilar 



Metals. 



The average resistance of a section of the experimental bar was about 

 0.00022 ohm. In the measurement of so small a resistance, and particu- 

 larly the change in this resistance on reversal of the heating current, it 

 was extremely important to avoid errors due to thermo-electric disturb- 

 ances at junctions of dissimilar metals. The potentiometer was specially 

 designed for the experiment and was made of 100 turns of manganine 

 wire wound on a marble cylinder. All connections to the manganine 

 were of copper and the instrument was completely enclosed in a glass 

 covered box. All adjustments of the contact point could be made with- 

 out opening the box. The resistance boxes A and B, Fig. 2, were also 

 enclosed in glass covered boxes. 



The fine potential leads, 3, 4, 5, etc., Fig. 2, were copper and were 

 copper-plated to the experimental bar. The connection with the poten- 

 tiometer was made through copper binding posts. The galvanometer key 

 and circuit were entirely of copper. 



It was found that copper and manganine form a thermo-electric couple 

 only about one eighth as strong as copper and German silver, hence the 

 choice of manganine for the potentiometer wire. 



The Heating Current. 



In the experiment on the change of resistance on reversal of the cur- 

 rent, it was very important that the current used for heating the wire 

 should be kept very nearly constant, since a change in it would cause a 

 change in the temperature of the bar and hence in the resistance of any 

 section. If t1iis change were confounded with the change due to the 

 Thomson Effect, serious error would result. The following arrangement 

 for measuring and regulating the current was found to be very satis- 

 factory. 



