Chemical Affinity in terms of Electromotive Force. 341 

 Maximum deviation 



from mean. 1.555-1-5003 

 On side of excess . 1.5003 =3*65 per cent. 



n m *a*- 1-5003-1-449 . AQ 



On side of deficiency i.*;n03 =o , 42 „ 



per cent. 

 " Probable error " of one determination ±0*0202 = ±1*34 

 „ „ result . . . ±0-0048= ±0-32 



Experiments on a possible Source of Error in Joule's Valuation 

 of J made in 1867 (Brit. Assoc. Reports). 



47. The method used by Joule in these experiments con- 

 sisted essentially in the comparison of the resistances of a 

 platinum -silver wire (made into a coil and immersed in a calo- 

 rimeter) and of copies of the B.A. unit, and then passing a 

 current, measured by a tangent-galvanometer, through the 

 coil and determining the amount of heat produced therein 

 in a given time. The mean temperature of the calorimeter 

 throughout a series of observations being determined, the re- 

 sistance which the wire would possess at that temperature was 

 calculated from the observed resistance at some other tempe- 

 rature, the rate of variation in resistance of the wire with tem- 

 perature used being known. Calling this calculated resist- 

 ance R, the average current being C, and the time t seconds, 



C 2 R£ 

 the value of J was deduced by the formula J=— =- ; where H 



is the heat evolved. 



It is evident that any error in the valuation of R must 



produce an error to exactly the same extent in the value of J 



thus deduced ; and that such a source of error must have 



existed is evident from the circumstance that inasmuch as 



heat was continuously developed in the interior of the wire, 



and as an interval of time must necessarily elapse before heat 



could pass from the interior of the wire into the water by 



conduction, the temperature of the interior of the wire, and 



consequently the mean temperature of the wire as a whole, 



must have been somewhat higher than the mean temperature 



of the calorimeter. Apart from this, too, the wire must have 



been heated to a small extent above the whole mass of water 



in the calorimeter, owing to the adhesion to the wire of a film 



of warmed water which could not possibly be displaced by 



stirring absolutely pari passu with its becoming warmed. The 



very efficient automatic continuous stirring-arrangement 



adopted by Joule in the third series of investigations detailed 



in the paper referred to must necessarily have reduced this 



source of superheating of the wire to a minimum ; but it is 



