346 HEJOIY A. EOWLAND 



on some system, and then express all our results in that system: the 

 accuracy of science demands that there should be no ambiguity on that 

 subject. In deciding among the three systems, we should be guided 

 by the following rules : 



1st. The system should be perfectly definite, so that the same tem- 

 perature should be indicated, whatever the thermometer. 



2d. The system should lead to the most simple laws in nature. 



Sir William Thomson's absolute system of thermometry, coinciding 

 with that based on the expansion of a perfect gas, satisfies these most 

 nearly. The mercurial thermometer is not definite unless the kind of 

 glass is given, and even then it may vary according to the way the bulb 

 is blown. The gas thermometer, unless the kind of gas is given, is not 

 definite. And, further, if the temperature as given by either of these 

 thermometers was introduced into the equations of thermo-dynamics, 

 the simplest of them would immediately become complicated. 



Throughout a small range of temperature, these systems agree more 

 or less completely, and it is the habit even with many eminent physi- 

 cists to regard them as coincident between the freezing and boiling 

 points of water. We shall see, however, that the difference between 

 them is of the highest importance in thermometry, especially where 

 differences of temperature are to be used. 



For these reasons I have reduced all my measures to the absolute 

 system. 



The relation between the absolute system and the system based on 

 the expansion of gases has been determined by Joule and Thomson 

 in their experiments on the flow of gases through porous plugs (Philo- 

 sophical Transactions for 1862, p. 579). Air was one of the most 

 important substances they experimented upon. 



To measure temperature on the absolute scale, we have thus only to 

 determine the temperature on the air thermometer, and then reduce 

 to the absolute scale. But as the air thermometer is very inconvenient 

 to use, it is generally more convenient to use a mercurial thermometer 

 which has been compared with the air thermometer. Also, for small 

 changes of temperature the air thermometer is not sufficiently sensi- 

 tive, and a mercurial thermometer is necessary for interpolation. I shall 

 occupy myself first with a careful study of the mercurial thermometer. 



(6.) The Mercurial Thermometer 



Of the two kinds of mercurial thermometers, the weight thermometer 

 is of little importance to our subject. I shall therefore confine myself 



