214 ROYAL SOCIETY OF CANADA 



vitiating the results, and I feel satisfied that the relative values of both 

 accidental and systematic errors are fairly reliable. 



So far then as stars of similar type to /? Orionis are concerned, it 

 is safe to conclude that the best results will be obtained by using a sift 

 about 0.051 mm. wide and that even with a slit up to 0.076 mm. wide 

 not much loss of accuracy will ensue. As the exposure time required for 

 these two slit widths is only about one-half and one-third that required 

 for a slit 0.035 mm. wide, it is evident that a considerable saving in time 

 can be effected with, at least, equally accurate measures. 



'There seems to be no reasonable doubt of this conclusion applying 

 to all early type stars, although it must not be forgotten that with wide 

 slits some of the faint metallic lines in hydrogen stars will become im- 

 measurable. However Avhere, as in our new single prism spectrograph, 

 the collimator focus is considerably longer, over 60 per cent in this case, 

 than the camera there will not be much loss in defining power for a slit 

 width of 0,051 mm., and this width, entailing a saving of one-half the 

 exposure time over a width of 0.025 mm., may safely be used. 



In solar type stars, if measured in the usual way, the question of 

 purity is a considerable factor and doubtless the narrower slits will give 

 more accurate values. If the spectra are measiired on the spectro- 

 comparator, however, it seems probable, as no question of identification 

 and the determination of the w^ave-lengths of blends can enter, that the 

 errors should be accidental errors of setting only and should be of the 

 Fame order as those determined above. This, however, can only certainly 

 be determined from a similar series of tests. 

 Dominion Observatory, Ottawa, May, 1900. 



