[PLASKETT] RADIAL VELOCITY DETERMINATION 87 
None of the orbits determined with low dispersion at Ottawa and 
Allegheny are of stars with solar type spectra so that no measure can be 
thus obtained of the relative accuracy of this dispersion. Three solar 
type binaries observed at the Lick Observatory with a single prism 
spectrograph show fairly accurate results especially W Sagittarii with a 
plate error of + 0.90 km. and which if three or four discrepant observa- 
tions are omitted reduces to + 0.55km. Thisis of the same order as 
the probable error of a single prism plate of Arcturus as determined 
here (+ 0.70 km.) 
The probable errors of single observations of binary and constant 
velocity solar type stars with three prism dispersion both at Lick and 
Ottawa seem to be very close to half a kilometre and in cases where it 
is greater it is apparently due to the poorer quality of the spectra for 
measurement. Three cases in which the probable error of a plate is less 
than half a kilometre are known to me. At Ottawa a series of 11 plates 
of 8 Geminorum gave a probable error of + 0.40 km. per second and 
when observations are limited to certain hour angles and special pre- 
cautions taken as at Bonn by Kustner in determination of solar parallax 
where 16 plates of Arcturus gave a probable error of asingle plate + 0.22 
km., and at the Royal Observatory Cape of Good Hope where 22 plates 
of 8 Geminorum gave a probable error of + 0.34 and 55 plates of œ 
Bootis of + 0.42 km. There is no doubt, though no values have been 
published, that the work at the Lick and Yerkes Observatories on solar 
type constant velocity stars is equally accurate. 
It seems to me therefore that we may safely draw the following 
conclusions from the preceding discussion :— 
I.—The accuracy of determination of the radial velocity of stars 
of solar type by means of spectrographs of different dispersions is not, as 
would be expected, inversely proportional to the dispersion but in the 
cases under discussion only a small increase of probable error, 40 per 
cent., takes place when the dispersion is divided by three. As the 
relative exposures required are as about five to one it is evident that 
stars more than a magnitude and a half fainter become available. 
II.—The probable error very rapidly increases with the increase in 
diffuseness of the lines in early type stars varying in low dispersion spec- 
trographs from about + 2to + 11km.persecond. Experience in work 
with these stars has convinced me that the whole of this error is not due 
to the accidental error of pointing but that in many cases some physical 
cause in the stars’ atmosphere is responsible for a considerable part of the 
discrepancy. 
III.—The result of this and other investigations shows that the pro- 
bable error of a single ordinary observation of a good second type star 
