230 
of the nearer stars, should greatly diminish his mean 
values. Among the classes in which the mean proper- 
motion is small, the percentage of exclusion is also 
small, and the results are but little modified. But it 
is noteworthy that the exclusion of 6 per cent. of the 
stars of Class K has reduced the mean proper-motion 
in a greater ratio than that of 28 per cent. of those 
of Class F, and also that the removal of one-fifth of 
the stars of Class G decreases the mean for the 
remainder to less than one-fifth of its initial value. 
It apvears from these results that a large majority of 
the stars of Classes F, G, and K have nearly, if not 
quite, as small parallaxes and proper-motions as those 
of Classes A and M, though they are not quite so 
remote as the stars of Class B. The large mean 
values obtained for all the stars of these classes are 
due to the presence of a relatively small proportion of 
near and apparently rapidly moving stars, of which 
the percentage decreases, but the mean proper-motion 
and parallax increase, from F to K. 
Campbell’s results * are derived from a comparison 
of the radial velocities and proper-motions of nearly 
1200 stars, mostly brighter than the fifth magnitude, 
and averaging about a magnitude brighter than Boss’s 
stars, which would lead us to expect that their mean 
parallaxes should be 40 or 50 per cent. greater. In 
his work, ‘‘a few stars having proper-motions ab- 
normally large for their classes were omitted in 
accordance with definitely set limits’? (which unfor- 
tunately are not described more specifically). |The 
approximate percentage of exctusion is given in the last 
column of the table. It appears on inspection that the 
differences between Campbell’s and Boss’s results for 
stars of Classes A, K, and M arise mainly from the 
greater brightness of Campbell’s stars; those for 
Classes F and G are due mainly to the different per- 
centages of exclusion, and that the only significant 
difference is that Campbell’s B stars, though averag- 
ing much brighter to the eye than Boss’s, havea 
slightly smaller mean parallax, and therefore must be, 
on the average, of greater real brightness. 
Closely allied with these investigations is the deter- 
mination of the mean peculiar velocity of the stars of 
each spectral class. The results of Boss and Campbell, 
reached almost simultaneously, and from quite in- 
dependent data—proper-motions in one case and radial 
velocities in the other—are in extraordinary agree- 
ment. The values found for the average component 
of motion in any arbitrary direction are (in kilometres 
per second) :— 
Spectrum B A F G K M 
Campbell 65 105 144 159 168 17-1 
Boss 6:3... 1025) 16:2. 5218:6 5 Dea) ear 
The rapid increase of the mean velocity from B to 
F is very remarkable. The slow further gain from 
F to M would attract little attention if it were not in 
the same direction. 
It should here be added that the phenomenon known 
as preferential motion, or * star-streaming ’’—the ex- 
cess of the average peculiar velocity of the stars in a 
certain direction above those in the perpendicular 
directions—is almost absent in Class B, very con- 
spicuous in Class A, and somewhat less so in the 
following classes, being partially concealed by the 
greater average magnitude of the velocities. ; 
Another notable difference between the various 
spectral classes may be found in the number of binary 
stars, both visual and spectroscopic, among them. 
We may distinguish two classes of visual double stars: 
binary stars for which orbits have been computed 
(with periods rarely exceeding two centuries), and 
physical pairs, the real connection of which is proved 
M4 Lick Observatory Bulletin, vol. vi., p. 134, IQIt. 
NO. 2322) \VOLN O2)| 
NATURE 
[APRIL 30, i914 
by common proper-motion, but the relative motions 
ot which are slow, and periods long—probably often 
thousands of years. The counts of the two classes 
here given are from a list prepared in the course of 
my work, and include all stars for which the neces- 
sary data could be obtained, including many stars 
for which unpublished observations of spectra have 
been generously furnished me from Harvard. For 
the spectroscopic binaries, Campbell’s counts have 
been taken from his catalogue of 1g1o.!° They in- 
clude all the systems the periods of which were then 
known, and are divided into two groups, one including 
all the periods of which are less than ten days, and 
also all those the periods of which, though not exactly 
known, are described as short; the other all the 
known periods exceeding ten days, and those which, 
though not precisely determined, are known to be 
long. 
TaBLe IV. 
Spectra Visual Physical Spectroscopic binaries 
binaries pairs Short period Long period 
B 0 52 33 15 
A 14 152 15 I4 
F 33 115 Tvs 9 
G 24 74 8 14 
K 12 G28 ls Ol sh gta 
M OWE Soc Tas en On here ® 
It appears that, in Campbell’s picturesque phrase, 
visual double stars of relatively short period ‘‘ abhor” 
Classes B and M, the greatest number being of Class 
F, with G a good second. Among the physical pairs, 
of long period, the most favoured class is A. Class B 
is abundantly represented, and Class M very sparingly. 
The percentage of stars which are found to be 
spectroscopic binaries is very probably greater among 
Classes B and A than lower down the list. As time 
goes on, indeed, more and more of the stars of these 
“later”? types are found to be spectroscopically double, 
but of long period; but among these classes the detec- 
tion of such systems, where the range of velocity is 
small, is much easier than among the stars of the first 
type, the iines of which are diffuse. In any case it is 
certain that short periods are almost confined to 
Classes B, A, and F, and are especially abundant in 
the first of these. |The few short-period stars of 
Class G which appear in the table are all Cepheid 
variables, most of which were selected for observation 
on this account, and would not otherwise have got 
into the list. 
Finally, we may note that, among variable stars, 
those of the eclipsing type, such as Algol or Beta 
Lyre, are for the most part of Classes A and B, 
though there are a number of Classes F and G, and 
one at least of Class K; that the Cepheid variables 
are almost all of classes F and G, with a few A’s and 
K’s; and that almost all the irregular variables, and 
all the variables of long period, are of Classes M or N. 
Stars of Class M the spectra of which show bright 
hydrogen: lines are without exception variable, and 
almost all the stars of Class N are also subject to 
changes in brightness. 
(To be continued.) 
UNIVERSITY AND EDUCATIONAL 
INTELLIGENCE, 
CaMBRIDGE.—Dr. James Ward, professor of mental 
philosophy and logic, has been nominated to represent 
the University on the occasion of the celebration at 
Oxford on June 10 of the seventh centenary of the 
birth of Roger Bacon, and Dr. Sorley, Knightbridge 
professor of moral philosophy, to represent the Univer- 
15 Lick Observatory Bulletin, vol. vi., p. 38, rgto. 
