660 



PROCEEDINGS OF THE AMERICAN ACADEMY. 



during the months of artificial heating of the laboratory they need 

 not be regarded. There was perceptible leakage from the magnet- 

 circuit into the galvanometer-circuit during June. Efforts were made 

 to eliminate errors from such disturbances by combining series of ob- 

 servations taken under a variety of conditions, but we cannot be sure 

 that these efforts were entirely successful. It is unlikely, however, 

 that they affected vitally the results at which we arrived. 



hLeXltf 



2000 



4000 GOOO 8000 



Figure 14. 



10000 12000 



Longitudinal Thermomagnetlc Temperature-Difference. — With the 

 magnetic field and the //// current as in Figure 3, we may look for an 

 increase or decrease of the temperature-gi'adient along the plate as the 

 result of magnetic action. This effect was looked for in Plate 1, but its 

 existence was not proved. The method used was substantially the 

 same as that illustrated in Figure 9 for determining a transverse tem- 

 perature-difference. That is, circuits (1) and (8), acting thermo- 

 electrically only, were set against each other, each having one coil of 

 the galvanometer. The great initial difference of temperature between 

 the junction (1) and the junction (3) made it necessary to place the 

 outer junctions of the (1) circuit and of the (3) circuit in baths at 

 different tem})eratures. 



We thought at one time, in April, 1910, that we had found a rather 



