ponent exhibit, as other stellar spectra do, a profusion of dark lines. These photographs being repeated at different dates, it was natural to compare them, and it was noticed that the lines sometimes appeared double and sometimes single. So striking a circumstance, of course, demanded closer investigation, and presently it appeared that this opening and closing of the lines was a periodical phenomenon. The interval between one maximum opening of the lines and the next was fifty-two days. If the star were a single object, then this phenomenon would be inexplicable. It was plain that the object could not be a single star; it must consist of a pair extremely close together, and in rapid revolution. The doubling of the lines will then be readily intelligible. When one of the components is moving towards us while the other is moving from us, all the lines belonging to one system are shifted one way, and all those belonging to the other system are shifted the other way, the effect on the spectrum being that the lines appear doubled. When the stars are moving perpendicularly to the line of sight, then their relative velocities towards the earth are equal, and the lines close up again. We thus at once learn the period of the revolution of the two components. The lines must open out twice in each circuit, and consequently we have as the first instalment of the numerical facts of the system that the period of its revolution is a hundred and four days.
It is, however, a peculiarity of the spectroscopic process that it provides us with a wealth of information on the subject. The amount by which the lines open when they separate admits of accurate measurement, and as this depends on the velocities, it follows that we obtain a deter-
N
