MECHANICAL EQUIVALENT OF HEAT. 137 



clearly due to the thermometry rather than to experi- 

 mental errors in the electrical standards. I extract the 

 following from my paper in the Transactions: "No change 

 in the value of the various units, or constants, involved 

 in this investigation could bring our results into absolute 

 agreement with those obtained by Rowland, since, owing 

 to the difference in the expressions for the temperature 

 coefficients of the specific heat of water, it is inevitable 

 that our conclusions should agree at some one tempera- 

 ture ; but must necessarily differ when expressed in terms 

 of a thermal unit at any other temperature, and thus 

 changes in the values of the units would only alter the 

 temperature of agreement. For example, Dr. Guil- 

 laume has pointed out to me that the experiments of 

 Commandant Defforges lead to the conclusion that the 

 value of g at Greenwich should be increased from 981*17 

 centims. to 98 ['24 centims. A similar correction would 

 slightly increase Rowland's value of J, and thus cause 

 our point of agreement to be about 12° C. instead of 

 n'5° C. Again, the value of g, assumed by Lord 

 Rayleigh [Phi/. Trans., A, 1884, p. 427), would have to 

 be slightly increased, and the resulting values of the 

 electro-chemical equivalent of silver, and of the absolute 

 electro-motive force of a Clark cell, would require modi- 

 fication, but the only result of any such change would, as 

 before, be to shift the temperature of agreement. 



" It is, therefore, evident that the difference is chiefly 

 due to errors in thermometry." 



This method of testing our electrical units could only 

 be satisfactory if a direct determination of the equivalent 

 by mechanical work was accomplished with the same 

 thermometers used under the same conditions as in my 

 late investigation ; harmony between the results would 

 then be a clear indication of the validity of our system 

 of electrical units, for the errors arising from faulty 

 thermometry would be of small importance and the con- 

 tinuation of the experiments over the same range would 

 avoid discrepancies which have their origin in erroneous 



conclusions as to the capacity for heat of water. 



10 



