Chemical Affinity in terms of Electromotive Force. 347 



the wire to weigh 10 grammes and to have the specific heat 

 0*04 (approximately that of an alloy of one part silver to 



2*3 



two of platinum), would raise its temperature -=-~ — k-ttu or 

 nearly 6° per minute. ; 10x0-04' 



It would therefore seem from this rough estimate that the 

 heat-development produced in the B.A.-unit coil by the cur- 

 rents used may have been sufficient to raise its temperature 

 perceptibly above that attributed to it ; while the same result 

 would not be produced to any thing like the same extent with 

 the experimental wire, owing to the latter being immersed in 

 water instead of solid paraffin. If it be assumed that this heat- 

 development sufficed to raise the average temperature of the 

 B.A.-unit coil during the observations 5° above that of the 

 experimental wire (an amount of heating not at all unlikely 

 to have occurred), the ultimate effect of this would be to 

 cause the underestimation of R, and consequently of J, by 

 5x0-031 = 0-15 per cent. 



54. Hence, finally, putting together this source of error and 

 that due to the superheating of the wire (increased by varnish- 

 ing), there is reason for supposing that the value of J obtained 

 by Joule in 1867 by the electric-current method may be at 

 least 0*5 per cent, too small ; that is, instead of giving a value 

 1'31 per cent, larger than that deduced from Joule's water- 

 friction experiments of 1850 and 1878 (Part I. § 34), the 

 corrected value would probably be at least 1*8 per cent, larger 

 than the water-friction value. 



The practical conclusion to be drawn from these experi- 

 ments, then, is that, in any determination in which it is necessary 

 to pass a current through a wire for any length of time, an 

 appreciable error, through increase in the resistance of the 

 wire, will be brought about by its becoming superheated above 

 the temperature of the medium in which it is placed when the 

 current exceeds a certain limit, the exact value of which ne- 

 cessarily depends on the conditions. This source of error will 

 affect determinations of E.M.F. made by such methods as those 

 used by Latimer Clark in the valuation of his standard cell 

 (Proc. Roy. Soc. xx. p. 444), and will similarly affect any 

 experiments on the determination of J by methods based on 

 the same principles as those involved in Joule's 1867 observa- 

 tions, unless the mode of experimenting be modified in such a 

 way as to eliminate this source of error, or at least to render it 

 negligible. We propose to examine certain such modifications 

 which have occurred to us as being likely to reduce this source 

 of error to an inconsiderable amount. 



2C2 



