888 PROCEEDINGS OF THE AMERICAN ACADEMY 



The average probable error of the five minima observed is 1.3 min- 

 utes, or about one third of that of /? Persei. This ratio, as has been 

 already stated, was to be expected, since the rate of variation of the 

 stars is about as three to one. The average deviations from the 

 ephemeris, after applying the constant correction of thirteen minutes, is 

 only 0.7 minutes. It becomes still less if we adopt another ephemeris, 

 as will be shown below. Clouds or twilight prevented observations 

 on both sides of the minimum on every night except on February 17. 

 Accordingly, from a complete observation of a minimum under favor- 

 able circumstances we may expect an error of but a few tenths of a 

 minute. 



The systematic difference between the observers is found by dividing 

 the algebraic sum of the residuals of each by their number : the 

 algebraic sum of 86 residuals by Mr. Searle is — 5.30; of 85 by 

 Mr. Wendell, —0.87 ; and of 102 by myself, +5.62. The corre- 

 sponding corrections are, — 0.06, — 0.01, and -1-0.06. As each of 

 these represent over a thousand settings, the differences are not prob- 

 ably due to accident. The excess of the computed probable error in 

 the eighth column of Table X. over that to be inferred from the resi- 

 duals in the seventh column is partlydue to the neglect of these differ- 

 ences. If applied to the observations, they would make them appear 

 more accordant. They would not probably sensibly affect the form 

 of light-curve or the times of minima, owing to the distribution of the 

 measures of each observer. 



The variation in light is given in Table XI,, which is derived from 

 the light-curve described above, after applying a correction of thirteen 

 minutes to the assumed minimum. The ratios of light are given for 

 every half-hour, expressed in differences of magnitude, in logarithms, 

 and in numbers, the full brightness being assumed as the unit in the 

 last two columns. 



Some interesting theoretical deductions may be drawn from tin's 

 light-curve. For about an hour and a half the light remains sensibly 

 constant at 0. 1 10, or about one ninth of its full intensity. This interval 

 is over one third of tliat during whicli tlie light is increasing or dimin- 

 ishing. If the variation in light is admitted to be due to a dark, eclips- 

 ing satellite, the diameter of the latter must be ^/l — 0.1iO = 0.943 

 of that of the star, in order to sufficiently reduce the light. A some- 

 what less diameter is possible if we admit that the star, like our sun, is 

 darker near tlie edges than in the centre. The effect of this is probably 

 slight, or it would show itself in other ways. The longest period of 

 uniform mininmm light would occur if the satellite produced a central 



