1159 



this difference at the boiling point the deviations which remain 

 between our investigation and tiiat of 137c/ are none of them more 

 than 0.01 of a degree and 0.02 of a degree. It seems that the deviation 

 at the boiling point consists of two parts, viz. tlie deviation of 0.02 of a 

 degree in the determination of the pressure of hydrogen with the 

 same reading of the platinum thermometer P^/ (or others which give 

 the same result) and another deviation also of 0.02 of a degree in 

 the calibration of the platinum thermometer. It is therefore quite 

 possible that accidentally in the case of the boiling point an unfavou- 

 rable concurrence of circumstances has produced a deviation which 

 is larger than the probable error. In our series of measurements a 

 deviation of 0.04 of a degree from the smooth curve appears to 

 occur at H^ ; this reading must therefore be looked upon as suspect 

 and as possibly containing an experimental error. 



Taking into account the observation made at the end of § 4 

 it seems as if in the meantime the observations VII and VIII of 

 Table I, the one by Kamerlingh Onnes and Keesom at 789,1 (the 

 value obtained for 760 by intra})olation is mainly based on this 

 observation) and B^ in Table III have to be rejected. A renewed 

 set of measurements over the whole range remains desirable in order 

 to arrive at the complete accuracy which the determination of the 

 temperatures allows us to reach at present. 



