DEFLECTION OF LIGHT — DYSON AND OTHEES. 163 
The mean residual without regard to sign is ± 0.21", from which 
the probable error of a determination of Aa? or Ay is ± 0.22". 
Star 7 is much the brightest. Stars 1, 6, 11, 13 are rather bright. 
Stars 2, 4, 10, 12 are fainter and more comfortable to measure. 
Stars 5 and 8 are very faint. Arcturus is on the plates but is much 
too bright to measure. No measures have been rejected. 
The determination of the deflection on the eclipse plates is based 
on the declinations (y), and the last column of Table XII shows 
that on the check plates the y comparisons are free from any serious 
sj'stematic error. 
Star 7 is of particular interest ; its position near the center of the 
field corresponds to that of Xi, Xj Tauri in the eclipse field, from 
which the greatest deflection is expected. The images (which are 
not quite round) have the same characteristic shape. Further, the 
brightness of No. 7 corresponds with but exaggerates the brightness 
of Xi Tauri, which is the brightest star in the eclipse field. It is there- 
fore a valuable check to find that its systematic error in declination is 
insignificant compared with the displacement (of the order of 1") 
afterwards found for v.^ and v.^ Tauri. 
The systematic errors in right ascension are larger (probably 
through imperfect driving of the clock). They may affect the dis- 
placement indirectly through the orientation constant, but with 
much reduced effect. Allowing for this reduction in importance 
there appears to be nothing to trouble about. 
The primary purpose of the check plates is thus fulfilled. They 
show that photographs of a check field of stars taken at Oxford and 
Principe show none of the displacements which are exhibited by the 
photographs of the eclipse field taken under precisely similar instru- 
mental conditions. The inference is that the displacements in the 
latter case can only be attributed to presence of the eclipsed sun in 
the field. 
33. We turn now to the differences of scale between Oxford and 
Principe, which are given by the plate constants «, h, cl, e determined 
from the measures. As determined these include the effects of dif- 
ferential refraction and aberration. The latter corrections were cal- 
culated for each plate by the usual formulae and applied so as to 
detennine the corrected jDlate constants a% h% d% e^ free from dif- 
ferential refraction and aberration. Due allowance was made for 
the change in the coefficient of refraction owing to the difference of 
barometer and temperature (about 40°) between Oxford and Prin- 
cipe. The results are as follows (in units of the fifth place of 
decimals) : 
