1 66 THE SCIENTIFIC PAPERS OF 



and then measured its resistance at various temperatures, with the 

 following results : 



Cold ], 000,000 units. 



12,000 

 8,000 



At intervals whilst red-hot . ' 



b,ooo ,, 



3,7<)0 



4,0<)0 



At white heat 3,700 



700 



650 

 At intervals, in a gas furnace 



intensely heated 



(550 

 55<) 

 500 



The resistance of the cylinder, when cold, returned to its 

 original value, and after repeated experiment, produced the same 

 results, whence it follows that the amount of error caused by con- 

 duction of the pipe-clay cylinder, is practically inappreciable until 

 a white heat has been reached : but that in measuring tempera- 

 tures exceeding a white heat, it is the tendency of the instrument 

 to indicate a slightly lower value than the true one. 



In order to avoid inaccuracy from this source, it is desirable to 

 expose the instrument to intense heat for three minutes only, on an 

 average, at the end of which time the observation should be taken. 

 This period of exposure will have sufficed to heat the protecting 

 capsule, and the platinum resistance wire, to within narrow limits 

 of the full temperature of the furnace, whilst it will have been 

 insufficient to penetrate and soften the pipe-clay cylinder. The 

 error caused by an invariable and insufficient period of exposure is, 

 moreover, proportional to the temperature, and can be determined 

 by experiment at a temperature below white heat. 



In adapting the resistance thermometer to the measurement of 

 high temperatures, a wide range of resistance is obtained, and it 

 is no longer necessary to determine these resistances with the same 

 precision as in measuring slight variations of ordinary tempera- 

 ture. In this case I dispense with the use of galvanometers and 

 substitute for the same an instrument which I propose to call a 



