( 347 ) 
the opinion that the radiation of the components of the doublet is 
accompanied by a rather high percentage of ordinary light. 
The brightness of grating spectra is only small compared with the 
intensity of the original, incident light. The amount of ordinary 
light contained in the total radiation of the doublet easily could 
remain below the limiting value necessary for perception, and not- 
withstanding its intensity compared with that of the spectral line 
be by no means small. * * 
In order to decide experimentally between the two possibilities, 
I have made some experiments on the completeness of the circular 
polarization. 
As regards intensity of light Michelson’s echelon spectroscope 
certainly surpasses all other spectroscopes of high resolving power. 
It therefore was the most suitable instrument for the projected ex¬ 
periments and satisfied a first condition I had to attend to. 
A second condition to be satisfied for the projected experiments 
is that the source of light be as intense as possible. 
I therefore for the present only investigated the spectra of sodium, 
mercury and thallium, which can be obtained with great intensity. 
These few elements moreover have the advantage of exhibiting 
several different types of magnetic separations. 
The methods used for the investigation of the circularly polarized 
light are given in §$ 3—5. Particulars concerning the sources of 
light are contained in § 6. The observations and conclusions con¬ 
cerning the chief subject of this paper form the end of this com¬ 
munication $$ 7—13. 
3. Verification of the circular analysers. 
In order to investigate the circularly polarized light of the com¬ 
ponents it is simplest to use either quarter-wave plates or Fresnel s 
rhomb. 
The * last contrivance has the advantage of suiting simultaneously 
the necessary conditions for a large part of the spectrum. The 
quarter-wave plates, however, can be selected only for a very limited 
part of the spectrum. They can easily be adjusted for the purpose 
in vieyv rather by splitting mica sheets. 
The interference colour exhibited, when the plate is placed between 
crossed Nicols, in parallel light, gives a measure of the retardation 
produced. 
As the estimation of this colour is, however, always a matter of 
some difficulty and if a somewhat great accuracy is desired, it is a 
