130 



SCIENCE 



[N. S. Vol. XLIX. No. 1258 



near maximum. Spectra of this ts^pe are 

 practically certain evidence of variability. 



Finally the novEe are distinguished by very 

 peculiar and characteristic spectra, which 

 undergo equally characteristic variations as 

 their light fades. A very few other stars 

 which show spectra of similar character are 

 also variable (v Carinse and the nuclei of 

 two variable nebulse). 



Our first generalization may therefore be 

 stated in the form: 



Regular periodic variability has apparently 

 little connection with the evolutionary stage 

 of a star's history, while variation of a 

 roughly periodic or non-periodic type ap- 

 pears to he intimately associated with partic- 

 ular stages of development. 



This is after all very much what might be 

 expected, for regular variation keeping time 

 accurately suggests a process regulated by 

 gravitation or rotation, and hence not nec- 

 essarily connected with any stage of evolu- 

 tion, while other forms of variation may well 

 arise from the physical state of a star, and 

 appear only at definite evolutionary stages. 



A second general property of variable stars 

 is that, with insignificant exceptions^ they are 

 objects of great luminosity, far exceeding the 

 sun. The eclipsing variables, with a few ex- 

 ceptions, average more than fifty tunes as 

 bright as the sun. The irregular variables, 

 and the long-period variables at maximum, 

 appear to be comparable in brightness with 

 other naked eye stars of class M, and hence 

 about a hundred times as bright as the sun. 

 The Cepheids are among the brightest stars 

 of which we know, ranging from 100 up to 

 perhaps 10,000 times the sun's brightness. As 

 for the ITovse, we know as yet but little, but 

 that little indicates that, at maximum, they 

 may be even brighter than the Cepheids. 



In general, it appears certain that almost 

 all variable stars are what Hertzsprung so 

 f elicitiously calls " giants." According to the 

 theory which the speaker has had a share in 

 advancing this would mean that variability, 

 while not confined to any one stage of a star's 



2 A few eelipsiiig variables, and the stars in the 

 nebula of Orion recently discussed by Shapley. 



evolution, is a characteristic of its early 

 life — not the fiickering of the dying flame of 

 age, but the exuberance of extravagant youth. 

 Hertzsprung's suggestion that the very faint 

 dwarf red stars, which, all agree, represent 

 the last observable stages of stellar history, 

 should be investigated for possible variability, 

 deserves however more attention than it has 

 so far apparently received. 



"With these preliminaries, let \is turn to a 

 rapid survey of the individual classes of va- 

 riables. Here it will be convenient to reverse 

 the Harvard order, and begin with the eclip- 

 sing variables. 



Typical variation of this sort is imme- 

 diately recognizable, since it consists in regu- 

 lar interruptions of otherwise almost constant 

 brightness. This behavior suggested to Good- 

 ricke, more than a century and a quarter ago, 

 that the obscurations of Algol were due to 

 partial eclipses by a huge dark planet, re- 

 volving around the luminous star in the 

 period of variation. As this hypothesis con- 

 siderably antedated the discovery that true 

 binary systems existed among the visual 

 double stars, it was of striking originality, and 

 it may fairly be claimed that the history of 

 binary stars begins with Algol. Unfortu- 

 nately, the very boldness of the hypothesis 

 led to its neglect for a full century, xmtil 

 Pickering revived it, and Vogel's spectro- 

 scopic study of Algol gave it striking con- 

 firmation; and it is only within the last ten 

 years that the study of eclipsing binaries has 

 really come into its own as a branch of double- 

 star astronomy coordinated with that of vis- 

 ual or spectroscopic systems. 



The eclipse theory of the variation of stars 

 of this type now stands on about as firm 

 foundations as anything in modem astron- 

 omy, being confirmed (1) by the precise repre- 

 sentation of numerous well observed light 

 curves (the irregularities whose presence was 

 previously suspected disappearing with im- 

 proved methods of observation) ; (2) by the 

 fact that every eclipsing variable which has 

 so far been studied speetrographically proves 

 to be a spectroscopic binai-y in which the 



