STUDIES IN STELLAR STATISTICS. II. 103 
4) The luminosities of the stars generally increase with the 
temperature. There is, however, probably a relatively large dispersion 
in the luminosities of the less hot stars (types K and M), whereas the 
dispersion in the luminosities of the hotter stars is rather small (compare 
Meddel. N:o 68). Especially it is to be expected that the stars of the 
spectral type B all have nearly the same luminosity. For this reason 
I have chosen these stars as the subject of this investigation, This 
choice is all the more appropriate as the great majority of these stars 
in the heavens are registered in the admirable Annals of the Harvard 
Observatory. The luminosity of these stars is, indeed, so great that 
a star of this type, situated — as far as can be concluded — at the 
limits of our stellar universe, is scarcely fainter than the 8" magni- 
tude. We are thus in the position to get, with the help of the B-stars, 
what might appropriately be called a skeleton image of the Milky Way. 
5) Taking all stars of the spectral type B together and using 
the known values of their proper motions and radial velocities, I ob- 
tained from stars brighter than the 5" magnitude a value of the para- 
meter R equal to 4,76 Siriometers. It was however found that this 
value was correlated to the apparent magnitude of the stars, the fainter 
stars giving a somewhat smaller value, the brighter stars a higher 
value of À. 
This anomaly was found to be due to the differenee in lumi- 
nosity characterizing the subclasses Bo, Bi, B», Bs and Bs belonging 
to the spectral type B and used in the Harvard classification of the 
spectra. Leaving smaller divergencies out of the question, the stars 
of the subclasses Bi and Be, which possess the greatest luminosity, 
have a value of R equal to 7.4 Sir., whereas the three remaining sub- 
classes 50, Bs and Bs (Bs and Be are not here considered) are charac- 
terized by a value of R equal to 3.3 Sir. We get for each subclass, 
substantially, the same value, whether we determine R from faint or 
bright stars. The only exception to be made is in the case of the 
bright stars of the subclass Bo, which seem to possess a greater lumi- 
nosity than the other stars of this type. I have found no plausible 
explanation of this anomaly. It may be connected with the spectral 
peculiarities of these stars and I have applied to M* PIcKERING regarding 
this point, The bad postal communications between Sweden and Ame- 
