The present paper emiibodies an attempt towards a physical explanation of the ordered gradationl in the spectra of stars-a. subject in which pioneerirng work was done by the late Sir Normlan Lockyer, but which was worked up with systematic thoroughness at the Harvard College Observatory, unlder the lead of the late Prof. E. C. Pickering and Miss A. J. Cannon.* Durinig this interval the spectra of mrore than 100,000 stars have been photographed, classified, and published with full details in the Henry Draper Memorial Catalogue. The milost noteworthy facts which have been brought to light from these monumental studies have thus been summnarised by H. N. Russell.t " The spectra of the stars show remarkably few radical differences in type. More than 99 per cent. of them fall into one or the other of the six great groups which during the classic work of the Harvard College Observatory were Tecognised as of fundamental importance, and received as designations, by the process of the survival of the fittest, the rather arbitrary letters B, A, F, G, K, M. That there should be so few types is noteworthy, but much lmlore remarkable is the fact that they form a continuous series. Every degree of gradation between the typical spectra denoted by B and A nmay be found in different stars, and the same is true to the end. of the series, a fact recognised in the familiar decimal classification, in which B5A, for exanmple, denotes a spectrum half-way between the typical examples B anid A. The series is not merely continuous, it is linear. There exists slight difference between the spectra of different stars of the same spectral class, such as Ao, but these relate to mninor details. Almost all the stars of the small outstanding minority fall inlto three other classes (or rather four), denoted by the letters P, 0, N, R. Of these, 0 undoubtedly precedes B at the head of the series, while R aid N, which grade one into the other, comrie probably at its other end, though in this case the transition stages are nlot clearly worked out." Russell is of opinion that the principal differenices in the stellar spectra arise in the main from variationis in a single ,physical variable in the stellar