208 
ASTRONOMY: H. SHAPLEY 
DISCOVERY OF EIGHT VARIABLE STELLAR SPECTRA 
By Harlow Shapley 
MOUNT WILSON SOLAR OBSERVATORY. CARNEGIE INSTITUTION OF WASHINGTON 
Received by the Academy, February 23, 1916 
In three earlier communications to the Proceedings various phases 
of the problem of Cepheid variation have been considered. ^ No further 
introduction to the present note is necessary beyond the statement 
that the periodic change in color, which appears to be typical of all 
Cepheids, has already been found to correspond to normal changes of 
spectral class for two cluster- type stars, RS Bootis and RR Lyrae, and 
for one variable of longer period, 5 Cephei. In order to test to what 
extent the inconstancy of spectrum is a general phenomenon of Ceph- 
eids, some 150 spectrograms of representative variables of this class 
have been made with the 10-inch portrait lens and objective prism. The 
periods of the stars investigated range from 9 hours to 27 days; many 
are those for which spectroscopic orbits have been computed; some are 
well-known naked eye variables, others are much fainter stars; for some, 
changes in color have been suspected from studies of the Hght curves, for 
others the maximum intensity of the spectrum has been observed to shift 
toward the blue upon the approach to maximum light. For none, how- 
ever, has it been suggested, so far as I know, that the spectrum changes 
periodically along the normal spectral series. 
Following the numerical method of classification recently described 
by Adams,2 change of spectrum of Cepheids is susceptible of easy 
detection, even when the variation is very small. For example, in the 
work with low dispersion on F-type spectra the relative intensity of the 
G band and the hydrogen line Hy is particularly capable of showing small 
changes of type. The results can be summarized briefly. 
The table contains data relative to the spectra of eleven stars (all 
that are now known to have variable spectra), and some information 
concerning their light fluctuations. The 
G I observed range of spectrum variation is in 
' ' nearly every case smaller than the proba- 
ble total range, as observations at the 
exact time of maximum and minimum 
FTG 1. SPECTRA OF 6 CEPHEI wcrc not made. The accompanying fig- 
NEAR MAXIMUM (ABOVE) AND f J b b 
MINIMUM OF LIGHT (enlarged five -Qrc coutaius two objectlvc-prism spcctra of 
5 Cephei, showing the conspicuous change, 
from minimum to maximum light, in the relative intensity of the spec- 
tral lines. The more detailed appearance with high dispersion of a 
