80 C. J. MERFIELD. 



These ideal conditions could not be expected. In the 

 method here adopted, it is assumed that the error of the 

 quantity ± r' n =f r' s , deduced from the tables, can be 

 neglected, this assumption is very near the truth, if the 

 observed zenith distances of a pair of stars do not differ to 

 any great extent. In this respect much care was taken 

 in the selection of the stars to be observed, so that the 

 zenith distances of a pair approached equality. 



In the Talcott method for the determination of latitude 

 the idea is to eliminate the refraction. To obtain the 

 refractions by the method here shown, the idea is to 

 eliminate the latitude from the equations. 



To obtain values of r from equations (3) and (4) it is 

 necessary to know the correct declinations of the observed 

 stars, together with the sum of the zenith distances of a 

 pair and the difference between the amount of refraction. 

 The declinations of the stars adopted for this work are 

 those of certain fundamental ones contained in Newcomb's 

 catalogue. The positions of these stars are reduced to an 

 absolute system and the values here used are considered 

 to be definitive. The sum of the zenith distances have 

 been obtained by observation with the meridian circle 

 instrument, the difference of the refractions being deduced 

 from a standard table. The tables here used are those 

 which form an Appendix to the "Greenwich Observations" 

 for the year 1853. These tables are constructed from 

 Bessel's Tdbulce Regiomontance; assuming that the reading 

 of the thermometer attached to the barometer is the same 

 as the external one, this assumption will seldom lead to 

 any sensible error. 



The method here outlined has some advantages, firstly 

 the complete elimination of the latitude and its variation ; 

 secondly the elimination of the nadir observations, since 



