49, 



servations, and an inspection of the result of each observation 

 will s-hew that it is impossible a mean of 60 observations can. 

 be aftected by a greater error of observation than a very 

 small fraction of a second- 

 It occurred that the mean of the observations made and 

 r^ad off in day -light might differ from the mean, of the ob- 

 servations made near midnight. It soon however was satis- 

 factorily ascertained that the differences could not arise froni! 

 this cause. 



, Secondly, the difference cannot arise from errors of divi- 

 sion, for in fact the same divisions are used as to the same, 

 star. The correction of the mean of the microscopes, obtained 

 by observations of different stars, which have been used to 

 deduce the observed zenith distance of a Lyrae, although, 

 affected by errors of division, occasions no error in the result,, 

 because care has been taken that the numbers of observations 

 East and West should be nearly equal. The zenith distances 

 corrected for the mean of the microscopes have been put 

 down merely to shew the consistency of the observations.. 

 The means of the zenith distances at each time of the year,, 

 depend only on the observations of a, Lyr» itself.. 



Thirdly, it cannot arise from uncertainty in the changes 

 of refraction. This star is too near the zenith for^any materiuK 

 uncertainty of this kind.* - 



* These observations have been calculated by Bradley's refractions. Had they been ? 

 calculated- by the French Tables, (which I have used for a, Aquilae and Arcturus) ; 

 the parallax would have been about two seconds. This alteration arises from the differ- - 

 VOL. XII.. I 



