how a better value can be derived. The uncertainty should not, however, be more 

 than this in the case of a station such as Melbourne. 



(7) Probable error of the Potsdam observations. 



Including the error due to variability of the pendulums, the probable error of 

 the value of the Potsdam mean works out to less than 2-OxlO~ 7 sec., which we will 

 therefore take as the maximum value. 



The final probable error of the determination of the time of swing at the field 

 stations we will denote by K, 



where K == v/(l) 2 +(2) 3 +(3) 2 + + - - + (7 



For convenience, these values are put down in Table LXXI, in units of the 7th 

 decimal place of a second. 



TABLE LXXI.* 



From this Table, it will be seen that the columns numbered (5) and (6) the errors 

 of clock rate and the variability of the pendulums and the columns headed 

 p m and (7) are those chiefly effective in the formation of the probable error of the whole 

 determination. Thus, the Cape Evans observations have a large probable error largely 

 due to the poorness of the observations for clock rate. In this respect, the Cape Evans 

 results are disappointing, for all the care bestowed on the technique of the observations 

 was of no avail. In the Antarctic, the unfortunate transit instrument was blamed for 

 many troubles often unjustly! ; but, at least, the burden of the blame for the high 

 value of K, it can in no way escape. 



THE FINAL RESULTS. 



Table LXXII is formed after applying the correction for lag to the Potsdam I 

 observations, and a further correction to both the Potsdam Series, in order to reduce 

 the observations in the cellar to the standard position on pillar No. 32 in the Pendelsaal. 

 This amounts to an addition of 3xlO~ 7 sec. to the observed time of swing. 



* The figures for Cliristchurch have not been evaluated in view of the discordant results by the 

 different pendulums. 



f The seamen used to declare that the setting up of the transit instrument was an unfailing sign of 

 approaching blizzards as, indeed, it usually was. 



90 



