[PLASKETT] RADIAL VELOCITY DETERMINATIONS 213 



PROBABLE ERROR. (First Series) 



The above suinmaiies of the probable errors at different slit widths 

 show some curious and unexpected results. If the spectra produced at 

 Ihe four slit widths are examined and compared under the microscope 

 the apparent quality for measurement very rapidly deteriorates as the 

 slit width is increased and one would expect, especially in dispersion I 

 and II of the original investigation where the focal lengths of camera 

 and collimator are equal, that the accidental errors w^ould be consider- 

 ably more than doubled for the wider slits and one would also certainly 

 expect the error to be least for the narrowest slit width. But, as will be 

 seen in the tables above, such is by no means the case. Taken generally 

 they show larger probable errors for slit 0.025 mm. than for slits 0.038 

 and 0.051 mm., a minimum in general for slit 0.051, and not a very 

 marked increase, except in the case of the probable error of a plate in 

 dispersion I of the original series, for a slit 0.076 mm. wide. 



I can offer no explanation of this curious and anomalous result, but 

 there seems to be no doubt of the reality and certainty of the conclusion. 

 The number of plates used, though not as numerous as would be desired, 

 was yet sufficiently large to prevent one or two abnormal plates from 



Sec. 111., 1909. 1.3. 



