OP ARTS AND SCIENCES. 83 



should say that in one week thermometers which had not heen heated 

 above 40° should be ready for use again, the error being then supposed 

 to be less than 1 in 4000, and this would be partially eliminated by 

 comparing with the air thermometer at the same intervals as the ther- 

 mometer is used, or at least heating to 40° one week before comparing 

 with the air thermometer. 



As stated before, when a thermometer is heated to a very high 

 point, its zero point is raised instead of lowered, and it seems probable 

 that at some higher point the direction of change is reversed again ; 

 for, after the instrument comes from the maker, the zero point con- 

 stantly rises until it may be 0°.G above the mark on the tube. This 

 gradual change is of no importance in my experiments, as I only 

 use differences of temperature, and also as it was almost inappreciable 

 in my thermometers. 



Another source of error in thermometers is that due to the pressure 

 on the bulb. In determining the freezing point, large errors may be 

 made, amounting to several hundredths of a degree, by the pressure of 

 pieces of ice. In my experiments, the zero point was determined in 

 ice, and then the thermometer was immersed in the water of the com- 

 parator at a depth of about 60 cm . The pressure of this water affected 

 the thermometer to the extent of about 0°.01, and a correction was 

 accordingly made. As differences of temperature were only needed, 

 no correction was made for variation in pressure of the air. 



It does not seem to me well to use thermometers with too small a 

 stem, as I have no doubt that they are subject to much greater irreg- 

 ularities than those with a coarse bore. For the capillary ai-tion 

 always exerts a pressure on the bulb. Hence, when the mercury rises, 

 the pressure is due to a rising meniscus which causes greater pressure 

 than the falling meniscus. Hence, an apparent friction of the mercu- 

 rial column. Also, the capillary constant of mercury seems to depend 

 on the electric potential of its surface, which may not be constant, 

 and would thus cause an irregularity. 



My own thermometers did not show any apparent action of this 

 kind, but Pfaundler and Plattner mention such an action, though they 

 give another reason for it. 



(c.) Eelation of the Mercurial and Air Thermometers. 



1. General and Historical Remarks. 



Since the time of Dulong and Petit, many experiments have been 

 made on the difference between the mercurial and the air thermometer, 



