STUDIES IN STELLAR Statistics. Il. 99 
The mean errors in 6, show that the obtained difference be- 
tween the dispersions of the velocities along the X-axis and along the 
Y-axis is rather uncertain, whereas the deviating value of o, is well 
established. 
The mean errors in the coefficients of correlation show that 
only for 5, a value deviating from zero is made probable. We may 
henee conclude that a slight dissymmetry in the velocity ellipsoid in 
the plane of the Milky Way is probable though uncertainly determi- 
ned’. It will suffice here to neglect this dissymmetry and put 
o = Go, — 1.934 + 0.110 
Ost 320 RL 
We find here the result of GYLLENBERG (Meddel. Ser. II N:o 13, 
P. 42) confirmed, that the stars of the spectraltype B are characterized 
by a velocity ellipsoid having the smallest axis directed against the 
pole of the Galaxy and the two axes in the XY-plane nearly of the 
same magnitude. GYLLENBERG finds, indeed, also a small difference 
between the mean velocities in the XY-plane and it is interesting to 
find that — though the mean errors in his investigations are still 
greater than ours — his values very closely agree with those found in’ 
this investigation. His axis of X is directed, approximately, against : 
ihe vertex and as this direction does not deviate considerably from 
the galactic Y-axis defined in § 18, his Z,, in Table XIII is to be com- 
pared with my value of e,, whereas his 2j, may be compared with 
my &. We thus obtain the following comparison: 
GYLLENBERG CHARLIER 
e, 2.00 2.04 
a, 1.86 1.83 
e 1.27 1.33 . 
The agreement is much closer than could be expected from the 
mean errors and from the difference in the methods employed. 
S 26. Velocities along and perpendicular to the radius vector (from 
the centre). It is to be expected, because of the symmetrical distri- 
! This dissymmetry would probably vanish if all B-stars could be taken into account. 
