402 
ASTRONOMY: H. N. RUSSELL 
stars of Kapteyn's Scorpius-Centaurus group^^ would be ideal objects, 
if they were not too far south. 
For fainter stars, eclipsing and Cepheid variables are available. Of 
the 90 eclipsing variables whose parallaxes were estimated by Russell 
and Shapley,^^ 69 are fainter than the eighth magnitude, and the 
mean parallax of these is 0';002, while only ten per cent exceed 0!004. 
For the Cepheids of similar brightness, the parallaxes estimated by 
Hertzsprung and Shapley^^ g^j-g gygn smaller. 
When once the systematic errors have been tracked to their source 
and eliminated, an extensive program of observation can be undertaken 
with security. Much duplication of observations is desirable, for it is 
obviously better that the parallax of a star should be determined from 
the mean of three or four short series of as many different observatories 
than by a series with a single instrument, however long and elaborate. 
Certain objects for which especially accurate parallaxes are desirable 
should be observed at as many places as possible. Examples are binary 
stars, stars differing in absolute magnitude from the bulk of those of 
the same spectral class, or from the values predicted by the spectro- 
scopic method, stars with exceptionally rapid motions in space, planetary 
nebulae, etc. Attempts to determine by direct observation the mean 
difference in parallax between classes of stars with small parallaxes (for 
example, those of the third and fourth magnitudes, taken as a whole) 
should, in my judgment, be deferred until the systematic errors have 
been thoroughly cleaned out. 
5. Knowledge of parallax leads at once to that of absolute magnitude, 
which, in the interest and importance of its systematic relations to 
other characteristics of the stars, stands second only to spectral type. 
(a) The relations between the two afford a very interesting study, 
which has led Hertzsprung^^ and RusselP^ to the recognition of the two 
series of 'giant' and * dwarf stars, coincident in class B, but gradually 
drawing apart among the redder stars until, as Adams' spectroscopic 
results have recently confirmed,^^ they are completely and widely sepa- 
rated in class M. If Russell's views are correct, the existence of 
these two series is the key to the problem of stellar evolution. In any 
case, their existence must be accounted for, and will be of importance 
in testing any theory. The securing of additional data, especially 
regarding the absolute magnitudes of individual giant stars, is much 
to be desired. It is of importance to determine not only the mean 
absolute magnitude of the giant and dwarf stars of each spectral class 
(whenever the two are separated) but the dispersion of the individual 
values about the mean. Only when the latter is known can the results 
