I9IS.] OBSERVATION OF ECLIPSING VARIABLES. 55 



secondary minima. The eclipses are now coming over forty minutes 

 earlier than they are predicted by the elements determined from the 

 original series of observations. This is very surprising in view of 

 the accuracy with which the elements were determined. These ele- 

 ments were determined mainly from a large number of well ob- 

 served minima grouped quite closely about two epochs about 700 

 periods apart. If the time of eclipse is fixed by the observations 

 within a half-miniute in each of these two regions, then the period is 

 known to one-tenth of a second. In cases where both branches of 

 the minimum were continuously observed it seems hardly possible to 

 change the observed time of any individual minimum by more than 

 one or two minutes, and still represent the observations closely. 



The photographic record of RT Persei, generously furnished me 

 by Professor Pickering, showed that at — 7,500 P, or lyYi years 

 before the zero epoch, the eclipses were about 100 minutes late. The 

 average period, then, is nearly a whole second shorter than the period 

 determined from my original series of observations. This shorter 

 period should have caused my observed minima to run off ten 

 minutes from the predictions during the interval of 700 periods. 

 This is intolerable. Beside my own observations, there are a good 

 many observations by Wendell and several by Graff available. Of 

 course a single estimated photographic magnitude is not nearly as 

 accurate for determining the time of minimum as a series of photo- 

 metric observations right through the eclipse. 



Making now the correction to the shorter, average period, I have 

 plotted the new residuals, and find that two periodic terms, one 

 running its course in 12,000 eclipse-periods, or 2'/y2 years, the 

 other in one-third this time, or 4,000 periods, with coefiicients of 

 twelve and five minutes respectively, going through their zero values 

 on the up grade together, fit the observations pretty well. The 

 smaller period is of the order of magnitude of that to be expected 

 from the revolution of the line of apsides caused by the observed 

 prolateness of homogeneous stars. The amplitude is of the size 

 to be expected from the smallest value of the eccentricity in accord 

 with the observed shift of secondary minimum. 



The important question, and one which is difficult to answer, is 

 whether the time of the secondary eclipse shifts in either of these 



