February 7, 1919] 



SCIENCE 



129 



stars may profitably bo followed by photo- 

 metric methods of precision, as in the visual 

 observations of Wendell and of Dugan and his 

 students, or by means of '" physical photo- 

 meters," such as the selenium cell, or the 

 photo-electric apparatus, the last of which es- 

 pecially, in the hands of Stebbins and of 

 Guthiiiek, has proved to be the most precise 

 of all ways of measuring starlight. 



When such high precision is not necessary, 

 good results may be secured by Argelander's 

 method of direct estimate of the brightness in 

 comparison with neighboring stars of known 

 magnitude — either directly with the telescope, 

 as Luizet and Roberts have done with notable 

 success, or upon photographs, as by Miss 

 Leavitt at Harvard. 



Here the great Harvard Library of photo- 

 graphs is a rich mine, worked to but a small 

 fraction of its capacity for lack of miners. 

 Estimates of the photographic brightness of 

 a few stars discussed at Princeton, show that 

 the probable error of a determination of 

 brightness from a single plate is about ±0.06 

 magnitudes. Material for thoroughly reliable 

 light curves of all regular variables which are 

 brighter than the eleventh magnitude at min- 

 imimi already exists in this great collection, 

 and it is urgently desirable that more workers 

 should be provided to make it available. 



Still higher accuracy can be obtained upon 

 plates taken specially for photometric pur- 

 poses, as has been shown by Plummer and 

 his associates. 



The majority of variable stars, however, are 

 far from exactly regular; and, when successive 

 maxima may differ in an unpredictable fash- 

 ion by a whole magnitude, there is clearly 

 little advantage in observing to hundredths. 



There are hundreds of such stars, most of 

 them observable with small telescopes over a 

 good part at least of their range — and to fol- 

 low them all would tax the resources of the 

 regular observatories severely. In this field 

 amateurs have been able to make what is 

 probably their most noteworthy contribution 

 in the whole range of present day scientific 

 activity. The American Association of Va- 

 riable Star Observers, and its older colleague, 



the Variable Star Section of the British 

 Astronomical Association, have organized this 

 amateur activity in a highly successful fash- 

 ion, and observations of these previotisly 

 rather neglected objects are pouring in at the 

 rate of many thousands a year — affording ma- 

 terial which will be of inestimable value in 

 the future. 



The observations of variable stars show 

 that almost all of them fall into some one 

 of five natural classes — to adopt the very con- 

 venient di^nsion devised at Harvard. In the 

 order there used these are: I., Temporary 

 Starx, or Novae; II., Variables of long period; 

 III., Irregular variables; IV., Short period 

 variahles, or Cepheids. including cluster va- 

 riables; and v., Eclipsing variables, including 

 the Agol and Beta Lyrae tj'pes. 



Variables of the last two classes are strictly 

 lieriodie, and notably regular in their changes ; 

 those of the first three classes are not. But 

 before discussing the separate classes in detail, 

 we may well consider some of the general 

 properties of variable stars of all kinds. 



First, it is noteworthy that, speaking at 

 large, there is as wide a range in the spectral 

 types of variable stars as in those of the stars 

 as a whole. Every one of the principal spec- 

 tral types, and almost every subdivision of 

 these types, is represented among variables. 

 It appears safe to conclude from this that 

 reliabilitj% per se, is not confined to any one 

 particular stage of stellar evolution. 



There is however a decided difference in 

 this respect between the various classes of va- 

 riables. Ecliijsing variables, though mostly of 

 classes A and B, are found as far down the 

 sequence as class K. Cepheid variables are 

 known through the whole range from B to M — 

 that is. in all the principal spectral classes. 



Irregular variables, on the other hand, are 

 almost all of classes M and N — and therefore 

 among the reddest of the stars. 



Long period variables also belong, without 

 exception, to these spectral classes, and the 

 great majority of them to the subclass Md, 

 showing bright hydrogen lines, at least when 



