( 474 ) 
Supposing that the asymmetry of the separation is entirely due to 
the shift of the central line towards the red, one should conclude 
from the communicated numbers that the displacement increases nearly 
linearly with the strength of field. This investigation was made 
with Rowranp’s grating, the principal object in view being to prove 
the existence of asymmetrical separations. I succeeded in this respect, 
but I think now I have overrated the accuracy of the extremely 
difficult determinations of the amount of the asymmetry. In fields of 
the order of 20000 gauss the asymmetry is 35 thousandth parts of 
an Angstrom unit, while the RowLanp grating used permits in the 
chosen, first order to resolve lines, the difference of whose wavelengths 
is 0.12 AU. hence with the field intensities mentioned we have 
to do with a quantity which is already four times smaller than the 
limit imposed by the resolving power. 
It is only because we have to do in determining the asymmetry 
with a difference of two quantities which are above the limit set 
by the resolving power, that there may be question of measurement. 
_ When we reach however the utmost limits of the method used 
then sources of error come to the front, which partly are caused by 
our mode of appreciation of the distance of two adjacent lines, partly 
are connected with particularities in the formation of images by 
gratings, not yet sufficiently understood. 
It is therefore undoubtedly to be preferred to use for the further 
investigation of the shift of the central line a method warranting 
greater resolving power. GMELIN in his investigation has used MICHELSON’s 
echelon grating, and it seems that he has largely succeeded by syste- 
matic procedure to interprete quantitatively the results given by this 
instrument. His result therefore possesses high probability and more- 
over is now supported by the theory given by Vorer *) in order to 
explain the large asymmetrical separations, a theory which assumes 
the existence of couplings between the electrons. 
I thought it however to be worth while to investigate the matter 
by a method independent of RowLanp’s and MiIcHELson’s apparatus. 
Fapry and Prrot’s method seemed most appropriate. The greater 
part of the measurements communicated in this paper have been 
obtained with a 5 m.m. étalon, already used on a former occasion. 
Some determinations were made with an étalon with distance-pieces 
of mvar as suggested by Fasry and Perrot in order to diminish the 
dependence upon temperature. It was constructed for me by JoBIN. 
1) Vorer. Magneto-optik. S. 261. 
