MECHANICAL EQUIVALENT OF HEAT. 133 



nitrogen thermometer of the Bureau International appears 

 to place the measurements, over this range at all events, 

 on a sure foundation. 



Since the publication of my paper in the Transactions, 

 Professor Callendar and I have conducted a laborious 

 series of comparisons between the mercury thermometer 

 and a new and beautiful form of the air thermometer in- 

 vented by himself. Its indications are independent of 

 changes in atmospheric pressure during the experiments ; 

 thus no observation of the length of a mercury column 

 is necessary, and one great cause of experimental error is 

 avoided. 



The results have now been published, 1 and although in 

 some respects unsatisfactory, they are sufficient tc clearly 

 establish a limit of range-error, not exceeding '003° C. of 

 the nitrogen thermometer over a range of 14 to 25 C. I 

 therefore venture on the statement that the evidence is in 

 favour of the greater accuracy of my temperature measure- 

 ments when compared with those of the observers I have 

 mentioned. 



Another great difficulty faces the experimenter, viz., 

 uncertainty as to the thermal unit. 



The condition of the science of calorimetry will continue 

 unsatisfactory until some definite conclusion is arrived at as 

 to the fundamental unit. The old definition — the amount 

 of heat required to raise unit mass of water from o° to i° 

 C. — dies hard in spite of its absurdity, and it is time that a 

 successor was appointed. I have suggested the following 

 definition, "The quantity of heat required to raise unit 

 mass of water through 1* C. of the air thermometer at 

 1 5° C," and I trust that this may be adopted. 2 Until this 

 matter is settled it is impossible to satisfactorily compare 

 the values of J obtained by different observers, for we 



1 Proceedings Royal Soc, January, 1894. 



2 1 am glad to see that Mr. Glazebrook in the excellent text-book on 

 Heat, which he has just brought out, adopts a rational range of temperature, 

 viz., 4 to 5 C. I still, however, maintain my preference for the range I 

 have suggested. 



