162 



A. TANAKADATE. 



as uncertain : undoubtedly tlie very crude nature of the result 

 is due both to noncentrobaric distribution, and probably more 

 to the slenderness of tlie data. 



Fig. 11. 



Fig. 11 is the reproduction of those arcs, from which we 

 see that, notwithstanding the widely scattered distribution of the 

 points of intersections, the arcs all pass through the region with- 

 in a few degrees of the mean co-ordinate, apparently showing the 

 feasibility of the assumption. 



The angular distances of the five points from this mean point 

 of intersection, or the values of ß are 



I. 



on or) 



II. 

 19.°7 



III. 



17.°G 



IV. 

 14.°."3 



V. 

 12.°3 



The point of no vertical variation is about 20.°ö by interpolation. 

 This excludes the possibility of the point l:)eing an antijiode by 

 the second of (14) which shows that there can be no such circle 

 within 01°-] of that point, so that the position of the image 

 must be given by either 



