90 



Dr. Schuster on a 



Hence the interval in the thermometer A as reduced by 

 Joule is larger than that measured on the Tonnelot instru- 

 ment by about one part in a thousand. 



Second Series. Twenty-two comparisons were made 

 between 14 and 20 , each comparison involving a consider- 

 able number of readings. Calling t B the interval as measured 

 on the Baudin thermometer 



t = <f B (i --00084) . . . (2) 



The relation between the Tonnelot and Baudin scale was 

 found to be 



t T =-t ( 1— -00089) . . . (3) 



Hence the second series would give 



tj = t T (i + -00005). • • • (4) 



making the Joule scale agree exactly with the scale of the 

 French hard glass thermometer. 



With respect to the difference in the results obtained in 

 the first and second series, it is to be observed that, owing to 

 the inequality of the divisions, we can hardly trust the Joule 

 thermometer to give us readings which for an interval of 

 io° are accurate to one part in a thousand. In the second 

 series an attempt was made to eliminate the errors of 

 division as much as possible by taking a great number of 

 readings in the same region, but not within the same two 

 divisions. The second series is more to be relied upon also for 

 the reason that it was made when considerable experience 

 had been gained in the comparison of thermometers. I 

 attach therefore much less importance to the second series, 

 and believe that for all practical purposes we must take the 

 scale value of Joule's thermometers to be identical with that 

 of the Tonnelot instrument. 



We may now reduce Joule's equivalent to the air 

 thermometer of the Bureau International des Poids et 

 Mesures, and then find 



