( 375 ) 



with the same filling of the thermometer yielded the following results: 

 (cf. Comm, N". 95« Tab. VI and N". 95^ Tab. I) ') 



July 7"S '05 — 139°.867 



Oct. 26^1', '05 — 139°.873 



July 6th, '05 — 217°.416 

 March 3'-d, '06 — 217°.424 



June 30'^, '06 — 182°.730 

 July Q^^, '06 — 182°.728 



For the deviation of the determinations on one day from the mean 

 of the determinations on the two days follows resp.: 



0°.003, 0°.004 and 0°.001 so mean 0°.0027, 

 which harmonizes very well with the mean error derived above for 

 a single observation ^), from which appears at the same time that 

 different systematic errors are excluded. This justifies at the same 

 time the supposition from which we started, that the error in the 

 resistance thermometer may be neglected. 



Determinations with different fillings agree very well. 



The determinations made on July 6^^^ '05 and March S'^, '06 

 with the thermometer filled with electrolytic hydrogen (see § 3) and 

 those made on June 30^^^ '06 with the thermometer filled with 

 distilled hydrogen, give: 



mean of July 6t\ '05 and March 3'^, '06 — 217°.420 



June 30^^ '06 5"50' — 217°.327 i 



6u 5' _ 217°.362 mean — 217°.345. 

 6"25' — 217°.347 ) 

 If the last temperature is reduced to the same resistance as the 

 first, we find — 217°.400, hence the difference of these values is 

 0°.020, from which, only one determination being made, we must 

 conclude, that also with regard to the filling systematic errors are 

 pretty well excluded down to — 217°. 



§ 8. Results. 



It appeal's from the foregoing that with our hydrogen thermometer 

 determinations of temperature, even at the lowest temperatures, 



1) The temperature for June 30'^ '06 given here differs slightly from that given 

 in Table I of Coram. N^. 95% though both refer to the same resistance. This diffe- 

 rence is due to the fact that in Gomra. N^. 95'' the result of one reading has 

 been used, and here the mean has been given of more readings. 



2) It gives namely for the probable error 0^.0029, so only a trifling difference 

 with the above. 



