THE DOMINION ASTRONOMICAL OBSERVATORY. 
In previous reports descriptions of the instruments in the Obser- 
vatory have been given, and the nature of the work done has been ex- 
plained. 
During the past year the work has followed the same general lines. 
There has been no important addition to the instrumental equipment, 
though several minor improvements to the existing apparatus have 
been made. 
In the Division of Astrophysics the determination of the velocities of 
stars in the line of sight has been the principal work. During the year 
ending March 31st last, the number of stellar photographs taken for 
this purpose averaged 76 per month. These are all measured here with 
special micrometric instruments, and the results deduced. Seven orbits 
of spectroscopic binaries were thus determined during the year, and seven 
more in part. 
In the use of the spectrograph in these observations a weakness 
of the spectra at the violet end was observed. This was traced to error 
in collimation of the correcting lens, the supports of which within the 
telescope tube are subject to a slight degree of flexure in different posi- 
tions of the telescope. The resulting collimation error has the effect of 
spreading out the star image on the slit unequally, as regards rays of 
different refrangibility. Since, during the exposure, the visible part of 
the star image is held on the slit, the visual part of the spectrum is 
intensified at the expense of the actinic. As it is impossible to prevent 
this flexure, the collimation has been made adjustable, and the best ad- 
justment for different positions of the telescope has been determined 
experimentally. 
An investigation of the effect of width of slit on the accuracy of line 
of sight determinations has shown that equal accuracy may be obtained 
with the slit one-twentieth of a millimetre in width, as with the usual 
slit of half that width. 
The accuracy with different dispersions has also been the subject of 
investigation, and it has been found that the probable error of the 
velocity determined does not increase as rapidly as the dispersion de- 
creases, but at a less rate. Thus dispersions of 10 and 33 tenth-metres 
to the millimetre gave velocities with probable errors of 0.5 and 0.7km. 
per second respectively. Against this slight decrease in accuracy there 
is the important advantage in using the lower dispersion, that the ex- 
posure need not be so long, and fainter stars may be observed. 
