68 SCIENCE PROGRESS. 



with No. 11,048 and with some other thermometers. I 

 had but one suitable micrometer scale, and this was used 

 when observing No. 1 1,048, which was divided to T Vth of a 

 degree. The probable limit of observational error in this 

 case was therefore only o'ooi°C. The other thermometers 

 were observed through ordinary reading telescopes, and as 

 it was impossible to estimate fractions of the spaces less 

 than ^, the probable limit of error of a single observation 

 was o # oo5°C. The actual uncorrected readings differed by 

 as much as o*4°C. in certain cases. The corrected results 

 oi the separate readings differed in no case by more than 

 0"Oio°C. and only in one case by more than o*oo6 u C, 

 which is about the probable observational error. The mean 

 result of the whole of the thirteen comparisons made be- 

 tween 11,402 and my standard (taken at regular intervals 

 from 12 to 2i°C.) gives an absolutely identical result of 

 17*1417 ! 



Only four comparisons were made between 1 1 ,403 and 

 my standard, the means being 25*197 and 25*200 respec- 

 tively. The exact correspondence in the former case must 

 be in part fortuitous, but results such as those are a won- 

 derful testimony to the accuracy of the tables of corrections 

 supplied by the Bureau, and an examination of the full table 

 of numbers obtained by us during the above comparisons 

 would, I think, convince the most sceptical of the validity 

 and necessity of the various corrections. 



As regards objection No. 2, if accuracy of such an order 

 is possible there can be no doubt of its desirability. For 

 example, the perfection of our methods of electrical measure- 

 ments is now so great that errors amounting to 0'oi°C. in 

 the estimation of the temperature of Clark cells and of 

 resistance coils, etc., become serious, and all efforts to trace, 

 by means of a mercury thermometer, the variation in such 

 quantities as the specific heat of water and other substances 

 with change of temperature, are hopeless, unless the previous 

 standardisation of the thermometer has been satisfactorily 

 accomplished. 



I do not wish in any way to disparage the standardisa- 

 tions performed at Kew, but it is evident that they do not 



