VARIABLE STARS 
215 
The agreement is not good, but there seems to be a simple explanation 
of this. The illuminated hemisphere is twice as bright as the dark hemi 
sphere so that the spectroscopic measurements of the secondary do not 
refer to the centre of the disc. Plaskett does not note any difference of 
type, so that the lines which he measured could presumably only be 
produced in the illuminated hemisphere. Since B 2 = -306a the displace 
ment may well amount to -08a and the radius of the relative orbit is to 
be increased in the ratio 100 to 92. This gives 
M Bol. Mag. (from R) Bol. Mag. (from M ) Residual 
S x 2-35 + 0-16 + 0-81 + 0-65 
S 2 1-20 + 3-00 + 3-91 + 0-91 
and the accordance is now tolerable. The faults which remain are probably 
due to the roughness of the data (photometric and spectrographic) for the 
secondary. To reduce the residual for S 1 to zero we must further increase 
M x /M 2 and the revised value of M x will then be about 3-0. The primary 
then falls on the central line of the main series, and the secondary is a 
giant star as in ¡3 Persei. 
We have had two instances where, by expurgating the less reliable 
data and trusting to the radiative theory, the primary has been brought 
on to the line of the main series; and I believe that other examples confirm 
this, though less definitely. We might perhaps take the risk of assuming 
in the treatment of eclipsing variables that one of the conditions to be 
fulfilled is that the central temperature of the primary is 40,000,000°. If 
it is true that eclipsing binaries (i.e. very close binaries) are only formed 
by stars of this class, the fact must have some fundamental significance 
which we cannot yet understand.