400 
ASTRONOMY: H. N. RUSSELL 
among Cepheids — the star being always redder at minimum than at 
maximum. More recent observations show that changes in the spectrum 
go hand in hand with the others. 
In the case of eclipsing variables, these changes arise from a difference 
in spectral type between the components, and it is found that stars 
separated by an interval less than their own diameters, and therefore 
very probably of the same origin and age, may have spectra differing 
as widely as those of Sirius and Arcturus.^^ Observations of such sys- 
tems, when the eclipse is total, provide the only direct method at 
present existing for studying the relations between spectral type, color 
index, surface brightness, and density, which are of fundamental im- 
portance. The determination of the spectral type of the fainter com- 
ponents of such systems, though often very difficult, on account of their 
extreme faintness, deserves special effort. 
(c) The concomitant variations in brightness, color, and spectrum, 
which Shapley^^ has shown to occur in every Cepheid variable that has 
been properly investigated, indicate very strongly that the proximate 
cause of the changes in all three is a periodic variation in the surface 
temperature of the stars. Shapley's suggestions^ that these differences 
in temperature arise from some sort of internal changes, perhaps of the 
nature of periodic oscillations in the radius, density, temperature, etc., 
appears to be the best which has been yet made; but there are still grave 
difficulties in explaining how such pulsations should in all cases pro- 
duce the very distinctive form of the light curve, with its rapid rise and 
slow fall, and still greater trouble in accounting for the variations in 
radial velocity, which show so remarkable a relation, both in amplitude 
and phase, to those in light. It is in fact still doubtful whether these 
stars are really binary systems or not. Intensive studies of a number 
of these variables, including the greatest practicable variety of represen- 
tative cases, would be well worth while. 
(d) Still less is known concerning the very numerous variables of long 
period, and the roughly periodic and irregular variables. In the obser- 
vation of their changes in brightness, amateur observers may obtain 
results of much value, and, under the admirable cooperative schemes 
organized by the American and British Associations of Variable Star 
Observers, they are at present furnishing a great mass of valuable in- 
formation. Very little is known regarding changes in the spectra of 
long-period variables, except that they often exist,^^ especially as re- 
gards the bright hydrogen lines which are usually present at maximum. 
Observations of the color indices of these variables are also much to be 
desired. Certain peculiar variables, such as R Coronae and SS Cygni, 
