Comparison of Thermometers. 89 



shows considerable inequalities which must be due to errors 

 of division ; these irregularities are visible with the naked 

 eye, two successive divisions differing occasionally by almost 

 ten per cent in length. This irregularity renders an accurate 

 comparison somewhat difficult. 



The comparisons were made in a horizontal trough, 

 containing a large quantity of water, the temperature of 

 which was kept slowly rising by means of an electric current. 

 It is not necessary to enter here into all the precautions 

 taken to secure accuracy, as they are all well known to those 

 accustomed to similar work. The readings having been 

 taken in a horizontal position, it was necessary to find 

 experimentally the pressure corrections necessary to reduce 

 the indications to the vertical. Two sets of observations 

 were made ; in the first of these Joule's instrument was 

 compared directly with a thermometer made by Tonnelot, 

 divided into tenths of a degree and carefully calibrated by 

 the Bureau International. By means of this instrument 

 Joule's scale value can be directly brought into relationship 

 with M. Chappuis' air thermometer. The second thermo- 

 meter used in the comparisons was made by Baudin,; it had 

 a more open scale than the Tonnelot, being divided into 

 fiftieths of a degree. The Baudin and Tonnelot had been 

 most carefully compared with each other by Mr. Gannon and 

 myself in the course of another investigation. 



First Series. The Tonnelot and Joule A were compared 

 at 19 different temperatures between f and 30 . The 

 observations between 13° and 22° were reduced by the 

 method of least squares with the following result. 



If tj means a temperature interval on the Joule thermo- 

 meter reduced by means of Joule's factor and t T the corres- 

 ponding interval on theTonnelot thermometer, all corrections 

 having been applied, it is found that 



*=* T (i--OO093) . . . (1) 



