Existence of an Atmosphere surrounding the Sun. 449 



sight best to select a pair not differing much in longitude, and 

 differing widely in latitude. I find such a series recorded in 

 my books for June 22, 1854, and the five subsequent days, 

 after which the southern spot underwent such a change of form 

 as renders the remaining positions unfit for our present purpose. 



If p,, p2 be the observed distances from centre, and p^, p^ 

 the observed position-angles, the mutual distances (f> result from 

 the well-known expression 



cos </> = cos p, . cos p2 + sin pj . sin p^. cos (Px—Pi)- 



The following Table gives these observed and computed angles, 

 uncorrected for solar refraction. 



The next Table gives the same angles corrected for solar re- 

 fraction on the hypothesis h=^, wi=l-010. 



On comparing the individual discoi'dances in the two tables, 

 we here perceive that correction for refrac- 

 tion affects the angles (f> by an almost con- 

 stant amount, and that, consequently, the 

 case taken for discussion is less suitable 

 than it appeared at first sight, and is, in 

 fact, of little or no use for the purpose. Still 

 I have retained it in this paper, from its 

 informing us of one important circumstance, 

 that the latitudes of spots receive nearly con- 

 stant corrections for solar refraction, and 

 consequently that elements of the plane of 

 the sun's rotation deduced from uncorrected 

 angles will be sensibly unaffected with error. 



Let us see how this arises. If P be the sun's pole, C the ap- 



