78 SCIENCE PROGRESS. 



such temperatures differs considerably from the results 

 obtained by exterpolation the platinum thermometer yet 

 supplies a standard to which all such measurements can be 

 referred. Thus, as far as regards range, the advantages 

 are entirely on the side of the platinum thermometer. It is 

 true that mercury thermometers can, with proper precau- 

 tions, be used over a very considerable range of tempera- 

 ture, especially those excellent instruments constructed by 

 Niehls of Berlin, which contain gas at a high pressure. 

 Their great drawback, as far as accuracy is concerned, is the 

 difficulty with regard to the stem temperature, a matter of 

 great importance in thermometers of this description. The 

 circumstances under which they are used almost invariably 

 prevent the complete immersion ; thus an estimation of the 

 mean stem temperature is a matter of great difficulty, and 

 however accurately they may have been graduated by the 

 makers, it is hopeless to expect accuracy of a higher order 

 than i° or 2° C. The same remarks apply to the potassium- 

 sodium alloy thermometers, and the latter I find are in addi- 

 tion very subject to changes in zero after exposure to high 

 temperatures. 



The cumulative weioht of the evidence I have sum- 

 marised in this section is, I think, great, and it is 

 discouraging to find that so little use has as yet been 

 made of it in the scientific world. In many cases manu- 

 facturers have shown themselves ready to take advantage 

 of so simple a method of estimating high temperatures, 

 but a platinum thermometer is, as yet, rarely to be found 

 in a physical laboratory. True, Messrs. Heycock and 

 Neville have applied the method with complete success 

 during their investigations into the melting" points of alloys 

 {5), and Professors Dewar and Fleming have used it to a 

 certain extent in their determinations of low temperatures 

 {12), although the last-named observers evidently dis- 

 trusted the relation given by the (d) formula and have 

 stated their results in the platinum temperature-scale. 

 Such examples of its application, however, are rare, and 

 in no case (except in Callendar's work and my own) do 

 I find any record of its use for the accurate determination 



