ON THE MECHANICAL EQUIVALENT OF HEAT 417 



stand above the air thermometer between and 100, and that the 

 average for the Fastre at 40 is about 0-1 C. Using the formula given 

 in Thermometry this would produce an error of about 3 parts in 1000 

 at 15 C., the temperature Joule used. 



The specific heat of copper which Joule uses, namely, -09515, is 

 undoubtedly too large. Using the value deduced from more recent 

 experiments in calculating the capacity of my calorimeter, -0922, 

 Joule's number would again be increased 13 parts in 10,000, so that 

 we have, 



Joule's value 423-9, water at 15-7 C. 



Eeduction to air thermometer -|-1'3 



Correction for specific heat of copper. . -f- -5 

 Correction to latitude of Baltimore. . . -f- -5 



426-2 



It does not seem improbable that this should be still further in- 

 creased, seeing that the reduction to the air thermometer is the smallest 

 admissible, as most other thermometers which I have measured give 

 greater correction, and some even more than three times as great as 

 the one here used, and would thus bring the value even as high as 429. 



One very serious defect in Joule's experiments is the small range 

 of temperature used, this being only about half a degree Fahrenheit, 

 or about six divisions on his thermometer. It would seem almost im- 

 possible to calibrate a thermometer so accurately that six divisions 

 should be accurate to one per cent, and it would certainly need a very 

 skillful observer to read to that degree of accuracy. Further, the same 

 thermometer " A " was used throughout the whole experiment with 

 water, and so the error of calibration was hardly eliminated, the tem- 

 perature of the water being nearly the same. In the experiment on 

 quicksilver another thermometer was used, and he then finds a higher 

 result, 424-7, which, reduced as above, gives 427-0 at Baltimore. 



The experiments on the friction of iron should be probably rejected 

 on account of the large and uncertain correction for the energy given 

 out in sound. 



The recent experiments of 1878 give a value of 772-55, which re- 

 duced gives at Baltimore 426-2, the same as the other experiment. 



The agreement of these reduced values with my value at the same 

 temperature, namely 427-3, is certainly very 'remarkable, and shows 

 what an accurate experimenter Joule must be to get with his simple 

 27 



