( 42« ) 



and llie i-esislaiicc lli(>riii()iii(M(>r and xarialioiis in the loiiii)eratiire 

 i'oclïicieiits (d' tlie i'osislance '). 



The (lifFei-eiiee ajjpears lo he hirjj;er Ihaii we slioiild expeel afler 

 the cxpei'ieiiee made with the liighcr teni|)emtiires. Il may l»e tliat 

 in the measui-emenl of May 5 '()(), the first nieasnrenient made 

 ill li(iind hych'ugcji, in the measurement of the fesi.stanee or the 

 reading of the liydiogen thei-mometer a svstematie error luis crept 

 in whicli escaped our attentioji. At any rate it \vill l)e necessary to 

 repeat the calibration at these lowest temperatures. 



The differences treated iji this section, in so far as they go beyond 

 the expected limit of accuracy, point ])artly to abnormal sources of 

 error, partly to errors which in future may be prevented (as for 

 instance by always measuring small resistances by means of the 

 differential galvanometer) and it is probable that when we avail 

 ourselves of the experience made we shall reach also for tempe- 

 ratures below — 217° an accuracy to (P. 02. 



§ 4. In the same way as Comm. N". 95*^^ § 7 the following obser- 

 vations, where two resistance thermometers were simultaneously 

 immersed in the same l)ath, allow us to judge of the accuracy with 

 which a temperature is iixed I)y a give]i i-esistance. 



With /V/ \\G have mad'> an adjustment h» a definite temperature 

 at which the resistaiice of /*///; was determined, then the temperature 

 was changed a little and again i-ead on /*ti and then the resistance 

 of Infill was determined and reduced to the first temperature. 



temp, on Pti — 87o.54, resistance P^// 103.950 £i~) 



new „ „ „ reduced — 87o.54, „ „ 103.959 difference 0.009 a 



or 0\014 

 temp, on Pti — 216°.65 17.379 



new „ „ „ reduced — 216°.65 17.385 difference 0.006 n 



or 0°.009 



') Although it is not excluded that here the variation of the temperature coeffi- 

 cients is larger than to — 217°, this can by no means explain the large deviation 

 because the wire had previously been carefully annealed. Moreover it is difficult 

 to assume that impui'ilies of the gas in the thermometer would be the cau.se, for 

 then we must accept that about 0.7 ",(, of air has been present in the gas, which 

 is rather impossible on account of the great carefulness observed when the 

 thermometer is tilled. 



~) For small differences in the calibrations of Ptm and Pti we refer to Goinm. 

 N". 99'' where also the zero's are given. The differences result from a more 

 accurate determination of the ratio between the arms of the Wheatstone l)ridge 

 and of the resistance of the conducting wires. In observations whicli were made 

 from 1905—1907 it appeared that the zeros had remained unchanged loJess than 



oTTaTa (^^'^P- ^^^'^ Comm." N". 99^). 



