The essential feature of the Draper classification is that it aims at classing together similar spectra, relying on general appearance, and not on the measurement of any one line or group of lines. This has the advantage of distributing the material in the most natural groups possible, and a disadvantage in that different observers may find it [Pg 191] difficult to be sure that their criteria are identically weighted. That the original aim was empirical and not theoretical is clear from the introduction to the first extensive list of spectra classified according to the Draper system:[499] “It was deemed best that the observer should place together all stars having similar spectra and thus form an arbitrary classification rather than be hampered by any preconceived theoretical ideas.” The present classification was the natural outcome of such a procedure. As A. Fowler has remarked,[500] “the Draper classification is based essentially on the observed spectral lines, and in reality may be regarded as independent of any other consideration whatsoever. Even if we did not know the origin of a single line in the stellar spectra, it is probable that we should have arrived at precisely the same order.” The descriptions that are contained in the preface to the Henry Draper Catalogue, and which have long been classical, were designed to describe the salient features of the groups that had been formed. It is only in a somewhat restricted sense that they constitute the criteria for those groups. The descriptions were compiled from the spectra of apparently bright stars of the classes involved, but the greater number of the spectra actually classified are taken with such short dispersion that all except the very strongest lines are difficult to distinguish, and are certainly not susceptible of accurate measurement. This fact should affect the standpoint of those who criticize the “multiple nature” of the Draper criteria. A portion of one of the plates used in the classification is herewith reproduced with no magnification. This photograph should make it apparent to anyone familiar with the use of spectra that the classification of stars is very largely a practical problem. Instead, then, of examining the possible merits of the best theoretical classification system, it appears to be more useful to examine the physical implication of the most representative classification that it has been found possible to make in practice. The fact that the Draper [Pg 192] system is so representative has been regarded as one of its great merits, and has rightly placed it in the authoritative position that it occupies.