18 Dr. W. S. Day on a Comparison of 



mixed the stirring was stopped, and the inner box imme- 

 diately closed. A micrometer reading of the mercury thread 

 on one of the thermometers was then taken quickly, the 

 telescope shifted quickly to the other thermometer, and a 

 reading taken of it. A second reading of both thermometers 

 was also taken. Of course the first readings were the most 

 reliable, but when there was no delay there was hardly ever 

 any appreciable change of temperature observed on reading 

 the second time, except sometimes at the extreme temperature 

 of 40°. The second reading was taken more as a check on 

 the first, to make sure there was no error in reading. Having 

 made this observation at a certain temperature, the inner 

 box was opened, the water stirred up again, and another 

 comparison made in the same way. This would be, of course, 

 at a slightly higher temperature. These readings at two 

 neighbouring temperatures were averaged and the average 

 regarded as one reading. When the temperature was higher 

 than that of the room, the only difference made was that 

 steam was passed into the heating-tube just before a reading- 

 was taken. The water was stirred, and the thermometers 

 were brought to the position in which they were read with a 

 rising meniscus, and not with a falling one. This was made 

 very sure by observing one of the thermometers through the 

 telescope while the heating and stirring was going on. 

 Having made two observations at neighbouring temperatures, 

 the top of the tank was taken off, the Tonnelot thermometer 

 removed, and its zero determined. The Tonnelot thermo- 

 meter was then replaced, the top of the tank put on, the 

 water heated, and a comparison made in the same way at 

 the next temperature. It was not possible conveniently to 

 make comparisons at many points in one day. The intention 

 was therefore to take observations roughly at 10°, 20°, 30°, 

 and 40°. This was only approximately carried out, as it 

 sometimes became convenient, for a good many reasons, to 

 use temperatures intermediate to these. 



7. COMPAKISONS MADE. 



A number of comparisons were made between the Tonnelot 

 thermometers taken in pairs, between the Tonnelot and 

 Baudin thermometers, and between the Baudin thermometers. 

 The results of these comparisons are given in Tables V. and 

 VII. The comparisons may seem to be rather few in 

 number, but it was not thought worth while to make any 

 more, considering all the circumstances. It must be remem- 

 bered that the readings of the thermometers were made 

 micrometrically, so that the errors in reading are exceedingly 



