It is within the stars that the first signs of band absorption appear. The band is seen in some stars, and a drop in the continuous spectrum of Vega around 4160 has been ascribed[456] to the cyanogen band headed at 4215. Similar “band” absorption can be traced in other stars of Class , and is even seen as early as . Identification of the cyanogen band headed at 3885, which always accompanies the 4125 band, would confirm the attribution to cyanogen, but the violet band does not seem to have been observed. The wings of H , which are often wide and conspicuous, render it difficult to trace anything of the nature of band absorption near 3885 for an star. (c) The Classification of Stars.—Several lines of evidence have indicated that the classes into which the stars have been divided are not physically homogeneous, and the problem of their classification is one of the future tasks of astrophysics. It is hoped that the writer can in the future make a more complete discussion of the question than is here desirable, and therefore the present treatment is to be considered merely suggestive and tentative. The material quoted is slight, and must be increased before conclusions can be justified. It has been suggested[457] that a one-dimensional arrangement will not suffice for the classification of the stars. The spectra have been classified, at least for the hotter types, by the strength of the and lines of Ca+. With the dispersion used in making the Henry Draper Catalogue, these lines constitute the conspicuous difference between one spectrum and another, and are the obvious criterion of class. If the spectra are classed by the strength of these lines alone, the classification is of course quite unambiguous, and for a one-dimensional sequence of spectra it would have been ideal. That the classes so formed are not homogeneous[458] indicates that some second variable must be described in a satisfactory classification, and that the strength of no one line could have been used with any greater success than that of the [Pg 169] and doublet. Further, practical difficulty in duplicating the classifications has been caused by the fact that the and lines are so far into the violet that they do not appear at all on many slit spectra, such as those used at Mount Wilson, and when other criteria are chosen for classification, it is likely that the results will deviate somewhat from those of the Henry Draper Catalogue.