The following note lias been made by Mr. Curtis on the method of reduction of the 

 observations : 



" I have reduced each set separately first, and tried to get a value for the azimuth 

 (neglecting collimation). Then I took the mean of the lot as the true azimuth, and 

 applied this to all the results (except those showing a very great discordance). It was 

 then usually possible to obtain a fairly reliable value for the collimation, which appears 

 to be fairly definite for each set, although liable to irregular disturbances. Finally, 

 I have taken the mean of the most concordant results, neglecting those showing large 

 deviations from the mean (frequently the case with high stars)." 



The most noticeable point in Table XX is the large number of star observations 

 which have been excluded in arriving at the mean result. Such a procedure could 

 hardly be recommended, did we not know that there were large instrumental errors 

 liable to occur with this instrument, and of almost unknown magnitude. The observa- 

 tions of July 1st are hopelessly inconsistent and have been neglected ; no collimation 

 correction could here be deduced from the observations. On the other three days, it 

 will be noted that the value of Dt deduced from i' Scorpii is always widely different 

 from that deduced from the other stars, and that a Orionis does not give very good 

 results. The inconsistency of the first may be explained by the fact that it is the single 

 high star which has been observed. The poor results from a Orionis are probably 

 due to the fact that this star has so low an altitude that it just skims the top of the 

 Barne Glacier to the north, and is subject to lateral refraction. In the same way, 

 <j> Sag. and a- Sag. also skim the slope of the small hill called Vane Hill, and may possibly 

 be subject to the same abnormal refraction. 



If these four stars are neglected in obtaining the mean values of Dt on these days, 

 there remains only a single observation of ft Orionis which shows serious discordance. 

 Unfortunately, however, if all four stars are neglected we are left on some days with 

 too few observations for the calculation of rate, and it seems wiser to throw out only 

 those observations which appear very inconsistent. 



Clock rate. It will be noticed that the actual observations give from 



June 30th to July 1st " S.C." losing 30 sec. daily. 



July 1st to July 2nd " S.C." losing 0-32 



July 2nd to July 3rd " S.C." gaining 0-02 



Owing to the uncertainties of the work, it has seemed advisable to use the same 

 clock rate throughout, and the mean of 0-20 sec. losing has been chosen. This corre- 

 sponds to a correction to the time of swing of the pendulums of 12 sec. x 10~ 7 . 



The temperature was measured by Thermometer No. 41204. The corrections, deduced 

 from the comparisons in May, 1910, and Oct., 1913, are taken as under : 

 At 20 C. thermometer reads 03 C. too low. 

 ,, 10 C. ,, ,, correctly. 



0C. 02 C. too high. 



+10 C. o-02C. too low. 



36 



