526 
the greater part. Wave-lengths and intensities of the influencing lines 
are taken from Rowranpb’s tables. 
As “effeetive’ we entered a line of intensity ¢ (Rowland scale) if 
its distance from the line under examination was d or less, corre- 
sponding values of 7 and d being: *) 
i d. 
0 and 1 0.1 A 
2, 3, 4 0.2 
561, 8 0.3 
9 and heigher 0.4 
In a few dubious cases the general aspect of a group of lines had 
to be taken into account in judging whether a certain line could 
reasonably be expected, from the point of view of the dispersion 
theory, to undergo onesided influence by its companions. Differences 
of opinion on such cases will always be possible; but every effort 
has been made to apply the test with the utmost care. For that 
purpose three observers have selected the lines, first independently, 
then after discussion in common. 
We found 26 lines having the effective companion on the red side, 
and 24 lines having it on the violet side. 
The diagram shows the result. We have first plotted, for succes- 
sive intervals of 100 A, the mean values of all limb-centre displace- 
ments measured by Apams; they lie on the zigzag line. The small 
crosses indicate the means for intervals of 300 Á *). The intermediate 
curve R gives the progressing of the mean limb-centre displacement 
with wave-length *). Finally the displacements of the 50 influenced 
lines were separately entered. The small circles refer to lines with 
companion on the red side, the full dots to lines with companion 
on the violet side. 
!) There is, of course, something arbitrary in this assumption; by altering a 
little the corresponding values of { and d the numerical results will also be slightly 
modified; but we have ascertained that the validity of the inference is not thereby 
affected. 
2) In drawing up the means we have omitted the 17 lines whose intensity is 
greater than 10, as such lines show abnormally small displacements owing to a 
cause which the dispersion theory reveals at once. 
5) An explanation of the anomaly occurring in this curve between A 5200 and 
A 6000 has been proposed by one of us (Cf. P. H. van Cirrert, inaugural 
dissertation p. 54, Utrecht 1919). 
