February 7, 1919] 



SCIENCE 



133 



hours (the cluster variables) and seven days 

 (the Cepheids proper). The two groups are 

 almost separated, there being very few stars 

 with iieriods about two days. 



3. Tlie range of variation is always small, 

 very rarely over 1.5 magnitudes. Photo- 

 metry of precision is revealing many cases 

 of Cepheid variation of very small range, the 

 most notable example being Polaris. 



4. The variation is continuous, and the rise 

 to maximum is usually much more rapid 

 than the fall to minimum, while the maximum 

 is usually more sharply accentuated than the 

 minimum. There are a few exceptions to the 

 last rule, and also a few stars in which the 

 rise and fall of brightness are about equally 

 rapid, but none in which the rise is decidedly 

 slower than the fall. The light curve is usually 

 smooth and flowing, and secondary maxima or 

 minima are rare, though they appear to be 

 authenticated in a few cases. The general 

 form of the light curve (which greatly re- 

 sembles that of the velocity curve of a spectro- 

 scopic binary) is characteristic, and makes it 

 easy to identify variation of this type. 



.5. The photographic range of variation is 

 always greater than the visual, the curves 

 being similar in shape, but the photographic 

 amplitude about 50 per cent, greater. These 

 stars are therefore much redder at minimum 

 than at maximum. 



6. The spectra of the stars of period less 

 than a day are almost all of class A, (P 

 Cephei, with a period of 5'' is of class B), 

 those of three or four days' period are of 

 class F; of 10 days' ijeriod of class G; while 

 those of longest period are of class K or M. 



7. The spectral class of these stars varies 

 with the light, being "earlier" at Tna.ximum 

 and " later " at minimum. As measured by 

 the hydrogen lines, the range of variation is 

 nearly a whole spectral class on the Harvard 

 scale : as measured by the metallic lines it is 

 much less. 



8. All these stars show variation of radial 

 velocity, with the period of the light changes. 

 The eix)ch of minimum radial velocity (most 

 rapid approach) coincides closely with that of 

 maximum light, and that of maximum radial 



velocity with minimum light — no exception 

 having yet been found among some twenty 

 stars. 



9. The range in radial velocity is nearly pro- 

 portional to tliat in visual magnitude, at the 

 rate of about 47 km. for one magnitude. 

 Hence the velocity curve can be roughly 

 drawn when the light curve is known. The 

 correspondence of the two, however, is not 

 precise. 



10. Most of the brighter of these stars be- 

 long to Miss Maury's ac class, i. e., have 

 spectral lines sharper than the average. Ac- 

 cording to Adams' recent criteria, the spectra 

 of these stars indicate very high luminosity. 



11. The Cepheids projier (of period greater 

 than two days) show a strong concentration 

 toward the galactic equator, and have very 

 small proper motions which nevertheless show 

 conspicuous evidence of the drift due to the 

 sun's motion in space. Their peculiar veloci- 

 ties are small, about 10 km./scc. 



12. Those variables of the " cluster type " 

 which are found in the skj' at large behave 

 very diflerently, showing a small galactic con- 

 centration. In the few cases so far studied 

 their proper motions are considerable, (though 

 they are faint stars) and their peculiar veloc- 

 ities very large. 



13. A large number of Cepheid variables 

 occur in the Small Magellanic Cloud, and a 

 very definite relation exists between their pe- 

 riods and photographic magnitudes, the stars 

 of longer period being the brighter. As the 

 stars in the Magellanic Cloud must all be at 

 very similar distances, this indicates that the 

 absolute magnitude of a Cepheid is a function 

 of the period. 



14. Among the numerous short period 

 Cepheids in the globular clusters those in any- 

 one cluster are of almost exactly the same 

 median brightness — differing from cluster to 

 cluster. It is therefore very probable that all 

 " cluster variables " of i>eriod less than a day, 

 are of the same mean absolute magnitude. 

 The few Cepheids of long period which occur 

 in globular clusters are considerably brighter. 



15. From the parallactic motions of the 

 Cepheid variables which occur in Boss's Pre- 



