296 PROCEEDINGS OF THE AMERICAN ACADEMY. 



variation of the specific heat of water with variation of temperature, this 

 specific heat being called 1 for every temperature used. This inaccuracy 

 will be referred to again. 



These figures show a considerable range of temperature, and from 

 them it should be possible to derive an approximate value at least of the 

 temperature coefficient of K. The numbers given in brackets, however, 

 will not be used for this purpose. The numbers for July 1, 2, and 8 

 exhibit great differences between Ki and iu, and also between the values 

 of K. On those days, and those only, the pair of copper-German silver 

 junctions, used to determine the change of temperature of the upper 

 stream, were covered with shellac melted on. The coating thus obtained 

 was too thick, so that the junctions did not take the temperature of the 

 water with sutficient readiness. There are in the table above given 

 other values of K obtained at temperatures not very different from 

 those at which these rejected values were found. The values obtained 

 for K ai all temperatures above 70° differ much among themselves; but 

 it hardly seems best to reject them altogether in the attempt to arrive 

 at an approximate value of the temperature coefficient of K. The great 

 variation observed among them was probably due to unsteadiness of 

 temperature of the water streams when very hot, or to possible impair- 

 ment by the hot water of the shellac coating on the copper-German 

 silver junctions. 



All the values of K not contained within the bracketed lines will be 

 used in some fashion in estimating the temperature coefficient ; but they 

 will be used in two divisions, one for May and June, the other for July 

 and August. The reason for this division is that on August 5 some of 

 the lines of wire leading from the copper coatings on the iron disk were 

 found to be out of condition. The pairs of wires affected were 1, 2, 11, 

 and 12, the other nine pairs remaining in good condition. When this 

 partial breakdown began it is impossible to determine ; June 30 was 

 the last date on which all the pairs of wires were known to be in good 

 order. It has been shown in the early part of this paper that each pair 

 of wires gave about the same effect as any other pair ; therefore, as all 

 were joined in multiple, the failure of a few of them should affect the 

 total current but little, the resistance of the remaining pairs being but a 

 small part of the total resistance of the circuit. The failing pairs lay, 

 one in the outermost circle, two in the next, and one in the next. It 

 appears, from a comparison of the values of K obtained near 21° and 

 near 39° in May and June with the values obtained near 28° in July 

 and August, that the impairment of the wires or some other unknown 



