[PLASKETT] RADIAL VELOCITY DETERMINATIONS 211 



dence was placed in the results obtained by it. The dispersions dis- 

 cussed here are : 



(1) N"ew single prism spectrograph, collimator 51 mm. aperture, 

 765 mm. focus, Brashear " Single Material " camera objective of 455 mm. 

 focus, linear diopersion at H;- 33.4 tenth-metres per millimetre. 



(2) Three prism form of old spectrograph, Zeiss "Tessar' camera 

 objective, 300 mm. focus; linear dispersion at H;- 17.5 tenth-metres per 

 millimetre. 



At each of these dispersions four slit widths were tested 0.025, 0.038, 

 0.051, 0.076 mm. Ten plates at each of these widths were made 

 and measured for dispersion (1) and six at each slit width for dis- 

 persion (2). In the original investigation seven or eight star lines were 

 measured on each plate, but, as it was found that lower probable errors 

 were obtained when the j^three best lines, X 4481.400, A 4471.676, 

 and X 4340.634 only were used, the measures were confined to these 

 lines in the present series. In addition four comparison lines were 

 measured on each plate. 



All measures were reduced to velocities by a modification of Hart- 

 mann's method,^ fully d scribed in my 1907-1908 report to the Chief 

 Astronomer. Each line was weighted during the measurement and the 

 velocity of the plate obtained from the weighted mean. The residual for 

 each line was obtained from this mean and these residuals as well as the 

 residuals from the plate velocities in a scries at one slit width were 

 treated in the manner to be now described to derive the relative accuracy 

 for different slit widths. 



As previously stated, wàth the star chosen, /5 Orionis, in which all 

 the lines are single, there can be no difficulty with blends or identifica- 

 tions, the question of purity does not enter directly and consequently 

 case (a) above may be omitted from consideration. Limiting ourselves 

 to the two sources (6) and(c), above, of error, it is evident that they 

 are in a sense entirely independent of each other. The former, that due 

 to the increased breadth of tlie lines with increased width of slit may be 

 evaluated without considering the latter, which refers only to the sys- 

 tematic displacements of the lines as a whole, by treating the residuals 

 from the lines on each plate in the series at one slit width, these residuals 

 being obtained in the manner described above. Thus at each slit width 

 in dispersion (1) there are ten plates and hence the residuals from 30 

 lines. From these 30 residual? there was obtained in the well-known way 

 the probable error of a line of average weight which was taken as being 

 the best value of the relative accuracy of measurement at that slit width. 



1 Astronomische Nachsichten 1.5.", 110. 1901. 



