1116 
which indicates that the limb—centre displacement of a line is 
augmented or diminished by as much as */, (on an average) of its 
normal amount if another line is near. 
This comparatively great influence can only be explained as arising 
from the modification which the neighbour line brings about in 
the refracting power of the medium. The phenomenon thus proves 
the efficiency of anomalous dispersion in the sun; it strongly suggests 
that limb—centre shifts in general may be chiefly conditioned by 
the shape of the dispersion curve of the gaseous mixture; and this 
inference again vindicates our hypothesis that the distribution of the 
light in Fraunhofer lines is dominated by anomalous dispersion. 
Taking for granted the validity of this interpretation of the solar 
spectrum we should expect, moreover, that in the spectrum of the 
centre of the sun’s disk the Fraunhofer lines will also be generally 
displaced towards the red with respect to the positions of their cores 
(which are determined by the values of the free periods on the sun), 
and that these shifts will be comparable in magnitude with the 
limb— centre displacements. 
There is strong reason, therefore, to ascribe the observed centre- 
are displacements perhaps wholly, but at any rate for a considerable 
part, to anomalous dispersion — an explanation, indeed, confirmed 
by the fact that the principal characteristics of these displacements 
are very similar to those of the limb—centre shifts *). 
If we now imagine the observed centre— arc displacements to be 
reduced by substracting from them the purely solar displacements 
of the centre lines with respect to their cores (as mentioned above), 
the remaining shifts — if any — will be so small, that the existence 
of a gravitational displacement of the solar line-cores with respect 
le) 
to the terrestrial arc lines (expected to amount to from 0,008 A to 
0,014 A in the visible spectrum) appears highly improbable. 
Let us finally try to express numerically with how much confi- 
dence we may assert that the observations really indicate the existence 
of a mutual influence of Fraunhofer lines. 
One might indeed suggest it to be an effect of mere chance that 
the black dots in our diagram average so much higher than the 
circlets. The probability of that supposed casual event can be calculated 
according to the rules of the theory of errors. 
From the equations (1), relating to the Mount Wilson measure- 
ments, it follows that a line having a companion on the violet side 
shows a mean relative departure 
‘) Cf. Junius and van Cittert. These Proceedings 23, 530 (1921). 
