44 PROCEEDINGS OP THE AMERICAN ACADEMY. 



thermometers also were placed in similar oil tubes, lest the direct action 

 on them of the water and the vapors might affect them more rapidly 

 than the junctions would be affected through the oil. 



With this arrangement another test of couples A, B, C, and D was 

 made November 18, with ice and water and steam, as on November 5. 

 This gave a result about 1 .5 per cent less than the one found by Mr. 

 Churchill for the same range of temperature. 



On December 2, a test of the same couples was made with a range of 

 temperature from 0° to 56.9° (vapor of boiling acetone) ; and the result 

 was a value about 2 per cent less than Mr. Churchill had found for the 

 same range. 



About the first of December, 1904, four new calibration couples were 

 made from the same kinds of wire that had been used before. In the 

 present case the german-silver wire was annealed, in pieces about 1.2 m. 

 long, by sending through each piece a current of about three amperes 

 (requiring about thirty-five volts) until the wire glowed brightly in a 

 darkened room. Each piece stretched a good deal during the heating and 

 was allowed to hang as a free catenary in the air ; but on cooling it re- 

 turned, apparently, to its original length. Soft solder was used in making 

 these new couples, which will be called A', B', C, and D', respectively. 

 These couples were mounted and were tested in oil, just as the couples 



A, B, C, and D had been mounted and tested at the last. On December 

 5, experiments with A', B', C,' and D' gave a result about 1.2 per cent 

 greater for the range from 0° to 56-7 J than Mr. Churchill had found for 

 the same range. On December 6, with a range from 0° to 100°, ex- 

 periments with the same junctions gave a result about 1.7 per cent 

 greater than Mr. Churchill's for the same range. 



Taken all together, the tests now described, made with couples A, 



B, C, D, A', B ! , C, and D', in oil, confirmed the values found by Mr. 

 Churchill ; and accordingly his values were used in reducing the obser- 

 vations made with the couples which he had made and had attached to 

 the main iron bars. 



Perhaps more attention has been given to this matter than the immedi- 

 ate object in view, the determination of the Thomson effect, required. 

 For, as the couples were used, an error of, for example, 1 per cent in 

 the reduction factor applied to their indications would make, to be sure, 

 an error of one per cent in the estimation of the amount of heat generated, 

 but it would make also an error of one per cent in the same direction in 

 the estimation of the interval of temperature between the junctions on 

 the bars. These two errors would balance each other, except in this par- 



