306 
For the other tin and antimony series I have obtained no line 
that was split up or only one. 
When we study the above results, we see that of the series for 
which it is possible to give an opinion in how far they follow 
Preston’s laws, only series XIII is not in conflict with these rules. 
In series VII the resolutions gradually increase; the first and the 
last splitting up lie farther apart than the limit of errors. Besides a 
quadruplet occurs there in the midst of the triplets. Even if these 
triplets were at bottom quadruplets, of which only the middle com- 
ponent was seen unsplit, which is very improbable, even then the 
amounts of the resolutions of the outer components would not be 
in agreement with each other. — The foregoing does not plead, in 
my opinion, in favour of v. Lonuizen’s results, though there remains 
a possibility that there are series that do not follow the rule of Preston. 
The efficacy of the Zenman-effects for the discrimination of spectrum 
series was very apparent when Runew and PascHeN found double 
lines in the spectra of Mg, Ca, Sr, Ba, which lines were changed 
in the magnetic field in the same way as double lines in principal 
and subordinate series in the spectra of the alkali-metals, while 
similar results were also found for some double lines from the 
Ra-spectrum. The hopes raised in 1902 and 1904 by the said 
researches with regard to the finding of series are accordingly not 
realized as far as the metals examined by me, are concerned. 
I have further also tried to find regularities myself. For this 
purpose and also in what precedes I have made use of a graphical 
representation, which seemed convenient to me. I have arranged the 
resolutions in every element according to their amount, and then 
plotted them on the same scale vertically under each other, as is 
usually done when resolutions are to be compared inter se. Two 
successive resolutions on the whole differing but little in amount, 
there arises a curve. A vertical part represents a number of equal 
splittings up. When the corresponding lines are indicated by the 
side of the resolution figure, we have at once a survey of all the 
lines that possibly belong to aseries. Compare the graphical representation. 
In this the resolution figures are only represented half, which, 
however, does not give rise to difficulties, as they are symmetrical. 
The components are not indicated by single lines; I have blackened 
the whole region, where they can be found according to the probable 
error. The vertical lines traced in the figure, are at distances = half 
the normal splitting up. A sloping straight part of the curve means 
that between two definite amounts of splitting up the resolutions 
are regularly distributed. When I examined this for the resolutions 
