98 BRIDGMAN. 



matter, as it often is, it can be seen that a regular succession of points 

 may be obtained apparently lying on a rising curve. To avoid error 

 from this effect each individual set-up demanded individual study, as 

 the way in which the wires snap about is an individual matter. The 

 true curve can be found by taking a larger number of readings than at 

 first I thought necessary, and by studying the possibilities of obtaining 

 successions of points on single straight lines under very slow changes 

 of pressure. In one or two cases, after the effect had been recognized, 

 it was found that the results originally obtained were in fact of the 

 wrong sign, and that on setting up the specimen again and carefully 

 analyzing the situation the correct result indubitably appeared of the 

 opposite sign. Examples of this will be mentioned later. 



Experimental Procedure. 



The procedure for either the radial or the longitudinal flow method 

 was essentially the same. The specimen was first attached to the 

 three terminal plug, connections to the interior end of the plug being 

 made with solder. The plug with the specimen in place was then 

 screwed into the pressure cylinder, and connections made to the outer 

 end of the plug by soldering. The temperature bath was then put in 

 place, and temperature adjusted to 30°, which was the universal 

 temperature of these experiments. (The accuracy was not great 

 enough to justify an attempt to determine the temperature coefficient 

 of the pressure coefficient, such as had been possible in the case of the 

 measurements of electrical resistance, for example.) 



The heating current was now turned on and equilibrium waited for, 

 as shown by the constancy of readings of the thermo-couple. An 

 interval of 20 or 30 minutes was necessary to reach initial equilibrium 

 after adjusting the temperature bath. This time was nearly all con- 

 sumed in acquiring equilibrium throughout the interior of the pressure 

 cylinder, and particularly between the outer and inner ends of the 

 three-terminal plug. The leads through the plug were of necessity of 

 steel in order to obtain sufficient strength, whereas the metal at either 

 end of the plug was copper. Hence any temperature inequality be- 

 tween different parts of the plug introduces spurious e.m.f.'s into the 

 circuit, so that it is essential to wait for complete equilibrium. As far 

 as the specimen itself and the heat flow through it from the heating 

 element is concerned, equilibrium was attained very quickly. If the 

 heating current was turned on or off after the bath and cylinder had 



