ASTRONOMY: H. SHAPLEY 
135 
linked together the cluster-type and the similar longer-period Cepheid 
variables, all of which show periodic shifts of the spectral lines, permitting, 
on the Doppler principle, of interpretation as the result of orbital motion 
in abnormal but improbable double-star systems. The recent work on 
the spectrum at Mount Wilson with a 10-inch photographic triplet and 
an objective prism shows that the spectrum varies between classes A 
and F in the same period as the variation of light and radial velocity. 
The star is, then, a periodic variable in at least three different senses. 
Every thirteen and a half hours, due to some cause as yet not fully 
understood, the light increases rapidly, more than doubling its intensity 
in less than two hours. During the same interval the apparent velocity 
of the star toward the earth increases from 47 to 91 kilometers a second, 
and the spectrum changes from type F to type A. 
This last variation is probably the most remarkable of all, for it im- 
plies, on the basis of our present knowledge of stellar radiation, a change 
in the temperature at the radiating surface of some three thousand 
degrees centigrade. Presumably the mean color of the light of RR Lyrae 
changes perceptibly during these cataclysmic two hours, but the evi- 
dence on that point is not yet conclusive. 
After the maximum of velocity, of light, and of spectrum have been 
reached, a reaction sets in, sharply at first, then more gradually, all 
factors decreasing for nearly half a day. The average time between 
outbursts is known to a fraction of a second, and the hour of a maximum 
a century from now could be predicted with considerable certainty. 
Doubtless a common cause underlies all these variations, and their 
proper interpretation and correlation will constitute the solution of 
Cepheid variation. 
Though variation of spectral type is suspected in the case of a few 
other cluster-t3^e variables, and in fact, on the basis of the present 
results, may be suspected with just cause of all short period Cepheids, 
the observations have definitely proved the spectrum variable for only 
two stars. For both of these the spectra have been studied at Mount 
Wilson, the data for RS Bootis being derived from spectrograms made by 
Pease at the principal focus of the 60-inch reflector. The results for 
the two variables are shown graptiically in the accompanying figure, 
where ordinates designate brightness in magnitudes and abscissae 
give time in hours. The light curve of RS Bootis, along which are plotted 
the determinations of the spectrum class for the corresponding phases 
of the light variation, is derived from unpublished photographic obser- 
vations by Mr. Seares and the writer. The Hght curve of RR Lyrae is 
derived from observations made at Harvard. 
