38 ART. 19. X. ICIIINOHE : 



(Allegli. Publ., T, 158) as follows; ''It is well known fact that 

 the spectra of Algol variables are almost always of very early 

 types. Our explanation for this is as follows : when a binary is 

 in its first stage the two components are of early type, their 

 diameters are large and their separation is small. Such a system 

 would appear as an eclipsing variable from all points situated in 

 a wide zone of the celestial sphere. As the type advances the 

 two stars become smaller and their distance apart increases. 

 Both these circumstances narrow the zone from which an eclipse 

 would now be visible. Consequently in any list of these variables 

 we should expect the number to decrease with the advance of 

 spectral type." This view seems to be acceptable in any way, 

 as it is difficult to think that the star become single as it ages, but 

 we ought not to forget to remark that even when we accept it, 

 still there are points difficult to explain. 



As we saw at the beginning of this article, Algol-type 

 variables are mostly found in the type A of the spectrum and 

 only six stars of the earlier type are included against 24 of the 

 type A. But when we pass to the spectroscopic binaries, the 

 phenomenon is somewhat different : looking over the lists given 

 in the paper by Campbell, above cited, we see that the number 

 of the spectroscopic binaries of the type O and B is 48, but 

 that of the type A is 28 so that, roughly speaking, it is the 

 half of the former. Why can we not discover the Algol 

 character in the earlier types in the equal proportion with the 

 spectroscopic binaries ? Will it be not easier to find out the 

 character or to see the eclipse for such earlier stars ? 



As we saw (pp 28-31), the stars which belong to the type 

 of Algol, tend to agglomerate in the plane of the milky way at 

 au extreme rate. Besides this, the ratio between the period 



