( 449 ) 
in our schemes has to be added to the irregular ones, contributes 
but R-light to the spectrum of the chromosphere. However, we are 
still perfectly ignorant, as to the proportion existing between the 
value of the radial gradient and the average value of the irregular 
gradients at a certain level. Considering the sharp contrasts in the 
complicated, well-defined structure revealed by the spectroheliograph, 
and leaving its interpretation aside, we cannot find it improbable 
that the local density gradients — whether in the mixture or in separate 
vapours constituting the solar atmosphere — should, as a rule, be 
steeper than the radial gradient. And if such is the case, the chances 
of V-light will not be much inferior to those of R-light. 
The wave-lengths of the most deflectable sorts of light on either 
side of a narrow absorption line differ only so little from the wave¬ 
length of the line itself, that practically a finite limit has not yet 
been assigned to that difference. In the spectrum those waves must 
therefore almost coincide with the absorption line. Displacements of 
the essential part, the “centre of gravity”, of a chromospheric line 
may of course occur, in consequence of the casual distribution of 
matter favouring at times R-light, then again V-light with the proper 
incurvation towards the earth ; but with the weaker lines such dis¬ 
placements are only slight, because they are restrained to the narrow 
region of great dispersion anomaly. 
The width of the wave-length region, in which a certain line of 
the solar spectrum produces appreciable anomalous dispersion, depends 
on the concentration with which the corresponding vibrating system 
is represented in the parts of the solar atmosphere traversed by the 
beam under consideration. With most lines that region is narrow. 
We obtain some idea of the width of it by noticing the widths of the 
Fraunhofer lines (wings included) of the average solar spectrum. This 
follows from our thesis that Fraunhofer lines are absorption lines 
enveloped in dispersion bands 1 ). According to Fabry and Buisson*) 
Fraunhofer lines of intensities 1 to 8 (on Rowland’s scale) have 
average widths ranging from 0,07 to 0,16 A, ; our theory requires 
♦he “centres of gravity” of chromospheric lines to be included within 
these regions (apart from exceptional cases relating to extraordinary 
local density gradients); so we cannot be surprised at the result, 
that the average difference between the wave-lengths of chromospheric 
and corresponding Fraunhofer lines is only ± 0.013 A. 
I see no reason, therefore, to conclude with Hale and Adams, that 
l > Pr °c. Roy. Acad. Amsterdam, XII, p. 280. 
a ) Fabry et Buisson, Gomptes rendus, 28 juin 1909. 
