70 
ASTRONOMY: H. SHAPLEY 
Proc. N. a. S. 
undertook an investigation of the periods and light curves of these stars, 
but for various reasons had not progressed far at the time of her death, 
beyond the work reported in Harvard Circular 173, 1912. In that paper 
she gives the periods of twenty-five stars, notes that they are all of the 
Cepheid type, and calls attention to a remarkable relation connecting the 
brightness of the variables and the length of their periods. The logarithm 
of the period was found to increase about 0.48 for each increase of one 
magnitude in brightness. The longest period she found was 127 days 
for one of the brightest stars in the Cloud; the shortest period was 1.25 
days for a variable nearly four magnitudes fainter. 
In the globular clusters are found large numbers of Cepheid variable 
stars with periods shorter than one day, and only a few with longer periods. 
In my work on cluster distances I used the relation given by Miss Leavitt, 
first putting it on an absolute luminosity basis (following Hertzsprung), 
and then extending it to these cluster type Cepheids. It soon appeared 
that the linear formula given by Miss Leavitt did not hold, and an empirical 
curve was derived and used for obtaining distances of Cepheids in the 
clusters and throughout the galactic system in general. 
Some objections have been raised to my extension of the period-luminos- 
ity curve, which identifies cluster type variables as giants rather than 
dwarfs. The proper motions of two of these variables are large, and many 
of them are in high galactic latitude. Both properties are usually, but not 
always, associated with low absolute magnitude. 
The adopted high luminosity is strongly supported, however, by the 
comparable brightness in clusters of the short period Cepheids and the B 
type stars, as well as by the direct comparison with the few long period 
Cepheids in Messier 5, Omega Centauri, and other globular clusters. 
Spectrophotometric analyses of the cluster stars afford additional support. 
A direct extension of the observed period-luminosity curve to cluster type 
variables in the Small Magellanic Cloud itself would appear to be especially 
decisive. 
Most of the known variable stars in the Cloud have median magnitudes 
corresponding to periods between two and five days. A few are fainter. 
About twenty-five of these faint variables have been investigated during 
the past year on a special series of long-exposure photographs made for 
Variable 
Period 
Median 
Variable 
Period 
Median 
d. 
d. 
1408 
0.715 
16.0 
1823 
0.795 
16.0 
1420 
0.67* 
16.4 
1928 
0.608 
15.8 
1634 
0.81 
15.6 
1943 
0.81* 
16.2 
1655 
0.444: 
16.2 
1964 
0.687* 
16.1 
1731 
0.65 
16.2 
2002 
0.702 
16.4 
1739 
0.63* 
16.4 
3610 
0.6 
16.2 
1741 
0.400: 
16.2 
