[PLASKETT] RADIAL VELOCITY DETERMINATION 85 
BP Orionis. I will in this case take the mean of the values obtained for 
the three slit widths 0.025, 0.038, and 0.051 mm. thus using 18 plates 
with III L, 30 plates with III R, and 30 plates with I. 
PROBABLE ERRORS. 



1 Line of gi 
Spectrograph. | Average Weight. ingle Plate. 
III L. 10.1 | 92.20 Er 
III R. 20.2 | + 2.47 + 1.53 
I 33.5 | +3.18 +2.16 

These results are, relatively to one another, in substantial agree- 
ment with those obtained in the present investigation though not quite so 
favorable to the low dispersion. There seems hence to be no reasonable 
doubt that the accuracy of radial velocity determinations does not by 
any means proportionately diminish with decrease of dispersion. 
It is evident therefore that the high probable errors of single obser- 
vations obtained in our work on spectroscopic binary orbits must be 
due, not to the small dispersion employed giving results relatively less 
accurate than those obtained with high dispersion instruments, but to 
the character of the lines in the spectra with the resultant high errors of 
measurement. In many cases also they are probably due to abnormal 
conditions in the orbit causing deviations from velocity curves due to 
simple elliptic motion, thus giving higher residuals. 
I have tabulated on following page the probable error of an average 
observation of the velocity of the brighter stars of spectroscopic binary 
systems as determined here and at the Allegheny and Lick Observatories 
and it will be noticed at once that the accuracy very rapidly diminishes as 
the spectrum lines become broader, and with orbits containing abnormal 
secondary or other effects. The slightly lower values obtained at 
Allegheny for the probable errors are likely due to the fact that many 
of their spectra were made on the fine grained Seed 23 instead of the 
coarser grained Seed 27 plate used at Ottawa. 
