272 BRIDGMAN. 



therefore, but I have taken the opportunity to record at least the 

 general nature of the results. 



The plan of presentation of the paper is this: first is given the 

 methods of measurement and computation, and then the detailed 

 account of the effect of hydrostatic pressure on thermal e.m.f ., Peltier 

 heat, and Thomson heat; this is followed by a brief presentation of 

 the effect of tension on thermo-electric quahty; finally the results are 

 collected for a general survey, and some theoretical deductions sug- 

 gested. This work was distinctly disappointing regarding theoretical 

 conclusions or suggestions as to mechanism. It appears from these 

 measurements that thermo-electric properties, as we measure them, 

 are probably a residual of different effects, sometimes working in 

 opposite directions, and that, for the present at least, these results 

 cannot be as suggestive as I had hoped for an electron theory of metals. 



In trying to interpret these results I have been obliged to think 

 through the general subject of thermo-electric phenomena, and have 

 been much impressed by the confusion that reigns in this field, even 

 respecting fundamental matters. As a preliminary to further work 

 I have brought together a number of general considerations, which 

 I hope may form a basis for at least consistent thought. These 

 general considerations seem proper subject for a separate paper, and 

 I hope to publish them elsewhere; they may, however, be read in 

 connection with this paper if the reader should at any time find him- 

 self in doubt as to my precise position. 



Apparatus and Experimental Methods. 



The general method is that which naturally suggests itself for meas- 

 uring an etfect as small as this, and is the same as that used previousl\' 

 by Wagner.^ Instead of using an ordinary thermo-couple, say of 

 copper-constantan, subjecting the whole couple to hydrostatic pres- 

 sure, and measuring the change in thermal e.m.f. at different pressures 

 when the difference of temperature of the terminals is predetermined, 

 the two branches of the couple are made of the same metal, the metal 

 under examination, and one branch only is subjected to pressure. 

 The metal under pressure behaves effectively like a diiferent metal 

 from that not under pressure, and we measure the thermal e.m.f. 

 of the metal against itself in the uncompressed state. The advantage 

 of this, of course, is that except for inhomogeneities in the wire the 

 entire effect measured is the effect sought, instead of only a small 

 change in a relatively large effect. 



