1904.] on Spectroscojnc Studies of Astrophysical Problems. \ 441 



and the major axis inclined, in the plane of the orbit, to our line of 

 sight, represented by the arrow-headed parallel lines. The line 

 Si So is the line of eclipses, or stellar conjunctions, and the two 

 lines of sight drawn normals to the curve are the two lines of 

 spectral-conjunction, when the star is moving in the tangent at right 

 angles to the sight-line. It is obvious that the excentricity of the 

 ellipse and the inclination of the axis may be such ac to leave no 

 measurable time-interval between the stellar and spectral conjunctions 

 at periastron, while the latter is considerably retarded at apastron. 



The importance of this retardation led me to look for some 

 independent confirmation of it, and I found it both in H^ and in H^, 

 but only in such manner as to show that with more photographs 

 clustering about the two epochs there might be confirmation amounting 

 to certainty. 



The line H^ has a dark line at a little distance on the more 

 refrangible side, which gives it the appearance of a double dark line 

 gradually closing up and reopening in the progress of the light 

 period. This appearance was found to be partly true by their 

 measured length-intervals, as may be seen in the tabulated results 

 projected on the screen. In this table the two dark lines are denoted 

 by Hg for the hydrogen, and c for its companion. There are 14 

 measures of the length-interval H^ -c in the first half of the cycle, 

 and 12 in the second half. All the plates were measured, but only 

 these could be measured accurately. On many plates the line c is 

 very weak, but these confirm the tabulated numbers inasmuch as all 

 belonging to the first half of the light period gave readings below the 

 general mean, and those of the second half gave readings above the 

 mean. The mean difference between the two sets of readings, treated 

 as a Doppler-efiect, would give a relative velocity of the origins at 

 about 100 km/sees. I have no doubt that Vogel's objection applies 

 here, and that the cause of the apjDarent shifts of the lines is the one- 

 side-thinning of dark H^ by its bright companion. This would 

 throw the apparent centre nearer to the fiducial line c in the first 

 half, and further away in the second half of the period. The range 

 of the readings could not be expected to give a reliable amplitude of 

 the oscillating displacement of the bright line, from which the 

 velocity could be deduced ; but the mean of the measures should give 

 the zero of displacement, and this is the reading of the plate belonging 

 to the periodic date 7-4. Unfortunately it is the only plate available 

 for the measurement between the dates 5-9 and 9-1 ; but on it the two 

 lines are very fine and clear, and its date separates all the lower 

 readings which precede it, from all the higher readings which follow 

 it. In this way the result is a striking confirmation of the one-day 

 retardation of the second spectral conjunction. It would be more 

 satisfactory if the coincidence of the first conjunctions had also been 

 indicated by the readings at the principal minimum. These indicate 

 correctly a preceding spectral conjunction, but occurring about eight 

 hours too soon. But on the plates at this epoch the two lines are 



