MECHANICAL EQUIVALENT OF HEAT. 



139 



Table II. 



As previously pointed out Joule's results are here given 

 as corrected after the comparison with Rowland's thermo- 

 meter, hence they have but little independent authority. 



Again, the mean temperature of Joule's temperature 

 ranges is difficult to estimate rightly, and 14 *6° C. can 

 only be regarded as an approximation. 



For purposes of reference and comparison it may be 

 convenient if I give a table of these values expressed in 

 terms of the different units usually employed. In each 

 case ^-=981 'iy or its equivalent in feet, the thermal unit is 

 that at 1 5° C. or 59° F., and the value at this temperature 

 is obtained by assuming the truth of my temperature 

 coefficient of the specific heat of water. The correction 

 thus introduced, however, is here very small, as will be 

 seen on comparing Col. I. in Table III. with Table II. 



Proc. Royal Soc, January, 1894, and they are due to the discovery of 

 an error in arithmetic. I have here made a further small correction, 

 amounting to 1 in 4000 only, which I did not feel at liberty to make in 

 the above communications, where I felt bound to adhere rigidly to the 

 actual numbers obtained by experiment. As I mentioned in the record 

 of the observations (p. 3S6), there was reason to believe that, although 

 the Clark cells used by me were for 'several days in the same bath as 

 the Cavendish standard, the temperature of the standard was slightly 

 higher, owing to its containing vessel projecting above the water. As it 

 was impossible to accurately ascertain the actual temperature within the 

 standard (no thermometer having been included in the cell) we did not 

 attempt to make any correction. On reflection, I feel convinced that it 

 is more probable that the cells differed in temperature by £° C. than that 

 they differed in E.M.F. As I am here estimating probabilities I have 

 assumed that this difference in temperature existed. 



