VI. 

 Discussion of Results. 



Let us first ascertain what is the probable error of a measured 

 coordinate, being careful not to let personalities in the observing 

 enter into our result. Each coordinate was measured completely, 

 that is in both the direct and in the reversed positions, hy two 

 observers. The difference between the two complete measure- 

 ments will be free from personalities and may be ascribed 

 to errors of observation. This difference, which I shall call v, 

 may easily be computed from Table III ; say the two observers 

 are Schlesinger and Kretz, then subtract (S — K) direct, from 

 (S — K) reversed and the difference is double the amount by which 

 one observer's complete measurement differs from the other's, or 

 2V. The probable error of a final coordinate is then given by, 



± 0.6745 /[ct] 

 2 ^ n 



Proceeding in this way for all the plates we obtain the follow- 

 ing probable errors. Onty those stars were used, thirty-three in 

 number, which appear on all the plates. 



Plate 





Probable Error of a 



Probable Error of a 





final x. 



final y. 



I 



±o r/ '034 



ic/'.C^I 



II 



.036 



.029 



III 



.023 



.023 



IV 



.024 



.020 



Y 



.020 



.027 



VII 



.034 



.020 



VIII 



.032 



.025 



IX 



•037 



.027 



Means, ±o // .030 zfco^.025 



The greater uncertainty in right ascension is due to the fact 

 that the images are usually elongated in that direction and 

 are therefore more difficult to bisect. The elongation was caused 

 by the failure of Rutherfurd's clock to keep pace exactly with 

 the diurnal motion of the group, sometimes lagging slightly or 

 sometimes moving too rapidly. 



272 



