394 
ASTRONOMY: H. N. RUSSELL 
(a) Two fundamental facts appear upon the study of the lines of stellar 
spectra. The first is that almost all of the thousands of lines which 
have been observed are identifiable as those of known elements, and 
can be reproduced under conditions which can be realized in terrestrial 
laboratories. The few outstanding exceptions are yielding year by year. 
The recent identification of the G band in the solar spectrum as due to 
hydrocarbons, 1 and of the bands of ammonia^ and water-vapor^ in the 
ultra-violet, leaves very few 'unknown' solar lines of any importance. 
Nor are there any of great account in stellar spectra, except in stars of 
the fourth type (Class N) and in the Wolf-Rayet and 'early' helium 
stars. 
So many of the lines in the latter have recently been found to be 
identical with those given in the laboratory by familiar elements (such 
as hydrogen, oxygen, carbon, and helium), under unusually intense 
electrical excitation^ that there is good reason to hope that further re- 
searches in this direction may account for those which still remain, 
and even solve the long-standing riddle of the origin of the characteris- 
tic nebular lines (which are associated with the Wolf-Rayet lines in 
the nuclei of planetary nebulae and in new stars at certain stages). 
The spectrum of the solar corona, however, still remains an isolated 
problem. 
(b) The second great fact is that the vast majority of stellar spectra 
fall into a single, continuous, linear sequence, which forms the basis of 
the Harvard system of classification, now generally adopted. Almost 
all the spectra which did not obviously belong to this sequence have been 
brought into connection with it by the recent work of Wright,^ con- 
necting the gaseous nebulae with the Wolf-Rayet stars at the head of 
the series, and that of Curtiss and Rufus,^ which shows that the small 
but definite classes R and N form a sort of side-chain, branching from 
the main sequence near the other end, at class G (or perhaps K) . Miss 
Cannon's experience^ in classifying over 200,000 spectra shows that the 
objects that do not fall into the sequence, thus extended, are almost 
vanishingly rare. 
The general characteristics of this sequence are now well established, 
and the types which were selected, by a sort of survival of the fittest, 
in the evolution of the Harvard classification prove to have been sur- 
prisingly well distributed along the series. With the aid of the quan- 
titative methods of classification developed by Adams and Kohl- 
schiitter,^ the precise classification of any spectrum of which a good 
photograph with suitable dispersion is available should be an easy 
