791 
of two lines seems the weaker, when it is in the neighbourhood 
of the minimum of the curve of intensity. These circumstances 
should be taken into account in determinations of the intensity with 
the echelon. This has not been done’) in former measurements’). 
The error caused by this, cannot be redressed by a small correction, 
but causes the relations of intensity found to be perfectly different 
from those that are present in incident light. 
The great importance of the function represented by (1) and fig. 1 
led us to test the theory by experiment before applying it to 
our measurements. For this purpose the intensities of the different 
orders of a spectrum line were measured (cf. fig. 1, spectrum line 
in four orders A, B, C, and D with intensities Aa, Bb, Cc, and Dd). 
Then the whole system of lines was slightly displaced by a small 
rotation of the echelon round an axis parallel to the effective sides, 
so that the lines assumed another position A’, B’, C’, D’, and the 
intensities A’a’, B’b’, C’c’, D’d’ were determined anew. When these 
measurements are repeated for some positions of the echelon, and 
when besides the position of every line with respect to a definite point 
1) The considered distribution of intensity has also influence on the observed 
position of the spectrum line when the centre of gravity or the maximum of inten- 
sity observed with the spectrum apparatus is understood by this. For a line within 
A 
the interval x) = — will be more greatly weakened on the outer side than on the 
o 
inner side, hence it will seem to be displaced towards the inside. A system of lines 
will, therefore, be compressed. The great divergency of the values, which different 
observers have found for the distances of the satellites of the green mercury line 
(cf. Nagaoka and Takamine, loc. cit.) is probably for the gredter part owing to this. 
8) NAGAOKA and TaKAMINE, Proc. of the Phys. Soc. of Londen 25, I, 1912. 
Tokyo Sûgaku-Buturigakkwai Kizi, 2e Serie, 7, I. 
