306 SIR F. W. DYSON, PROF. A. S. EDDINGTON AND MR. C. DAVIDSON ON A 

 21. The values of are collected in Table VII : 



TABLE VII. 



By subtracting the a of the comparison plates the scale plate is eliminated, and we 

 derive from right ascensions a = +0 r -105 and from declinations = +-0 r -098. 



Reference to the normal equations shows that the declination result is of double 

 the weight of that from the right ascensions. 



Thus 



- +0"-625. 



This is at a distance 50' from the sun's centre. At the time of the eclipse the sun's 

 radius was 15' -8 ; thus the deflection at the limb is l"-98. 



The range in the values of a is attributable to the errors inherent to the star images of 

 the different plates, and cannot be reduced by further measurement. The mean values 

 +0 r -015 and O r -031 arise from the errors in the intermediary scale plate. 



22. The probable error of the result judging from the accordance of the separate 

 determinations is about 6 per cent. It is desirable to consider carefully the possibility 

 of systematic error. The eclipse and comparison photographs were taken under 

 precisely similar instrumental conditions, but there is the difference that the eclipse 

 photographs were taken on the day of May 29, and the comparison photographs on 

 nights between July 14 and July 18. A very satisfactory feature of the photographs 

 is the essential similarity of the star images on the two sets of photographs. 



The satisfactory accordance of the eclipse and comparison plates is shown by a study 

 of the plate constants. The following corrections for differential refraction and 

 aberration are calculated from the times and dates of exposure. 



