634 PROCEEDINGS OF THE AMERICAN ACADEMY. 



The distance between the streams was about 6 cm., so that the grad- 

 ient of temperature in the wire between must have been about 10° per 

 cm. The ends of the wire, which was about 130 cm. long, were 

 soldered to copper wires, and the junctions were placed in a suitable 

 water bath, the copper wires leading to the galvanometer. The part 

 of the iron wire lying between the streams of hot and cold water was 

 placed across the space between the magnetic poles in a field the 

 strength of which, for a distance of perhaps 4 cm., was not far from 

 8000. 



Galvanometer readings taken with "field off" and "field on" 

 showed that the change of thermo-electric quality produced in the 

 unequally heated constantan wire by magnetic action, if any such 

 change occurred, was probably too small to yield a change of 3 X 10"^ 

 volt in the thermo-electric force of the circuit. 



Four couples of copper and annealed constantan wires were tested 

 for calibration purposes. The temperatures used were those of the 

 tap-water (about 12.5°), boiling ether (about 35.5°), boiling alcohol 

 (about 80°), and boiling water. The method used was substantially 

 that which has been described in previous papers."* None of the junc- 

 tions came into direct contact with any one of the liquids or its vapor, 

 each being protected by an arrangement like that shown in Figure 10. 

 The thermometers used had been studied with considerable care in 

 previous work. 



The four couples agreed so closely that for no one of the tempera- 

 ture intervals used in the calibration did the result from any one 

 couple differ so much as one per cent from the mean result given by 

 all. This mean follows : 



Temperature interval. Mean e. m. f. per degree. 



11.6°-34.5° 3911 absolute 



34.6°-79.1° 4145 



79.3°-99.9° 4527 " 



The thermo-electric current runs from constantan to copper at the 

 hot junction ; that is, at ordinary temperatures the copper line on 

 the thermo-electric diagram lies between the iroi! line and the con- 

 stantan line. 



Plotting a curve through the three points given by the data shown 

 above, we get from it the following table for copper-constantan : — 



* These Proceedings, 41, p. 42. 



