566 PROCEEDINGS OP THE AMERICAN ACADEMY. 



When we are testing our couples with junction temperatures of, let us 

 say, 140° and 180°, respectively, the german-silver wire between the 

 vapor pots being freely exposed to the air or contained in glass tubes 

 freely exposed to the air, we must have a very steep gradient of temper- 

 ature in those parts of the wire whicb lie near the entrance to the pots. 

 If near one pot there happens to be a section of german-silver wire 3 or 

 4 centimeters long having within itself a range of, let us say, 80°, and 

 if the thermo-electric quality of this section differs from the mean 

 thermo-electric quality of the wire one half as much as certain con- 

 tiguous sections differ in the wire which gave the data for Figure 8, the 

 electromotive force of the circuit as a whole, when the difference of 

 temperature between the (copper) -(german-silver) junctions is 40°, will 

 be about £ per cent greater or less than it would be if the given section 

 were of average quality. If this differential effect assists the current 

 when the neighboring (copper)-(german-silver) junction is the hotter 

 one, it will oppose the current when this junction is the cooler one, 

 thus making the difference between the " parallel " arrangement current 

 and the " crossed " arrangement current § per cent. If there is, as 

 there may happen to be, near the other pot an equally eccentric section 

 of wire, its eccentricity being of the opposite sign, we may get a 

 difference between " parallel " and " crossed " amounting to £ per cent. 



This goes some distance toward accounting for such difference (p — c) 

 as we have observed, but it goes hardly far enough. The mean of the 

 arithmetical values of (p — c) with the temperature interval 140° to 

 180° is about 1.3 per cent for the couples A 2 , B 2 , C 2 , D 2 , and, if we 

 confine ourselves to observations thus far mentioned, about 3 per cent 

 for the couples A 3 , B 3 , C 3 , D 3 . With couple C 3 on September 26 (p — c) 

 was more than 5 per cent of the mean effect. It seems doubtful 

 whether so great a discrepancy can reasonably be accounted for by such 

 heterogeneity as exists in a freshly annealed wire; but the results of 

 certain later calibration tests, now to be described, tend strongly to 

 confirm the opinion that the (p — c) value is due in some way to the 

 condition of the german-silver wire between the pots. 



Further Calibration Tests of Couples A 3 , B 3 , C 3 , and D 3 . 



We made on October 28 another calibration test with couples A 3 , B 3 , 

 C 3 , and D 3 , in which we tried to keep the upper parts of the glass 

 tubes containing the german-silver wires in a state of temperature 

 intermediate between the temperatures of the two pots, about 100° and 

 218°, respectively. With this purpose we closed in the air space about 



