(579 ) 
The additional lens being removed, the light-arrows forthwith 
re-appeared above and below the rather broad dark bands in the 
curved spectrum. 
It appears therefore, from our observations that in spite of the 
considerable width of the dark bands in the main spectrum, the 
corresponding light is but very slightly absorbed by the sodium lines. 
The flame has allowed every kind of light to pass, even that of which 
the wavelength differed ever so little from that of the D-lines; but 
it has caused these rays to be deflected from the straight line much 
more forcibly than the other parts of the spectrum lying further 
removed from the absorption lines. 
Here, then, we have a case where the absorption spectrum of a 
vapour exhibits broad bands not deserving the name of absorption 
bands. The special manner in which the experiment was made, enabled 
us to see what had become of the light that had disappeared around 
the sodium lines; but very likely the broad bands would have been 
attributed entirely to absorption if somehow this abnormally refracted 
light had fallen outside the field of vision of the spectroscope. 
In studying the absorption spectra of gases and vapours, we should 
be careful to see — which is not always done — that the absorb- 
ing layer shall have equal density in all its parts and shall not 
act anywhere as a prism. It would be worth while investigating 
in how far the anomalous dispersion can have influenced cases in 
which broadening or reversal of absorption-lines have been observed. 
In my arrangement the absorption-lines were narrow, if the main 
light had passed through a pretty much homogeneous and non- 
prismatic part of the flame. 
The experiment, as described above, offers no opportunity for ob- 
taining reliable values of the refractive indices. A better method 
to arrive at more reliable results is now being investigated; for the 
present all we can say is that the deviation of rays whose wave- 
length is very near 4p, or Ap, is at least six or eight times greater 
than that which the remoter parts of the spectrum were subject to. 
BECQUEREL says that the index for waves greater than Ap, and Ap, 
may attain 1.0009; for waves on the other side of the absorption 
line the index falls considerably below unity. The line D, produces 
in a much higher degree than PD, refractive indices smaller than 
unity '); the very high indices are represented in pretty much the 
same degree near D, and Ds. 
') In the woodeut Fig. 3, pag. 577 the upper arrow near JD, is spoiled and 
rather short compared with that near D. 
