( 672 ) 



with the observed coiTections to Souillart's theory and their probable 

 errors/) The residuals are verv satisfactory, especially so if satellite 

 I is left out of account. (See also Gron. Pahl. 17, pages 92 

 and 115). 



From the values of w^ in 1900, 1836 and 1750 I have already 

 in Gron. Publ. 17 derived the motion of to^. The value found there 

 requires however a small correction. The values which Bessel, and 

 following his example Schur also, gives for E^ sin 52^ and E^ cos 52^, 

 i.e. for h^ and k^, are in reality the values oï e ^ sin oï^ a,nd e ^ cos lo^. 

 This was not noticed at first, and must now be corrected. 



I now find for 1836 



h, = — 0°.704 k^ = - 0°.395. 



4 4 



Using, as before, the most probable values of e^, Wj and t^^, we 

 find from this : 



We have now 



from which 



^' r= 0°.001872 ± 0°.000020 (a) 



dt 



If the probable error were derived from the residuals, or from 

 the probable errors for the separate epochs, we should iind a much 

 smaller value. The larger value has been adopted chiefly on account 

 of the possibility of systematic errors of Bessel, which will be men- 

 tioned below. 



CooKSON has already (Cape XII. 2. page 197) derived the motion 

 of w, from the observations of 1836, 1879 (Schur) 1891, 1901 and 

 1902. He finds: 



^ =i0°. 001892 ± 0° 000024, (/?) 



dt 



The values («) and (^i) agree within their probable errors. So, 

 if (,:?) were adopted instead of («)> the final results could only be 



') in deriving these residuals the longitudes of the perijoves are, of course, counted 

 from the point 0, as was done in the tabular places. 



