When a group of stars is being studied for a special purpose, it is often found that the Draper classes are not fine enough to subdivide the material usefully. In such cases reclassification is often essential. It has sometimes been suggested that this indicates that the Draper classes are inadequate; but it must be recollected that, for the greater part of the material contained in the Catalogue, finer classification would have been impossible, and the subclasses in use today represent the practical survival from a far larger number, which were originally thought to be usable. Actually the stars represent a continuous gradation from class to class, and in classifying it is only possible to use the smallest distinguishable steps, which will obviously be smaller, the larger the dispersion. When it is found necessary to reclassify the stars more finely in a special investigation, as in the Harvard or Mount Wilson work on spectroscopic parallaxes,[501] one or more measurable criteria are selected and used as a basis, but standard stars classified at Harvard are used to define the scale. These measured or closer classifications, while essential for the purpose for which they were designed, have no theoretical advantage over the Draper system (on which they are ultimately founded), and do not, as is sometimes inferred,[502] indicate that the latter is in error. Although devised with no theoretical basis, the Draper classification has long been recognized as classifying something physical, and the fact that the majority of the stars had been ranged by it in a single sequence suggested that a single variable was principally involved. From general theoretical considerations it could have been predicted that this variable was probably the temperature, but, in addition, the observational evidence that this was the case was immediately convincing. In the words of A. Fowler,[503] “... the typical stars not only increase in redness in passing through the sequence, but successive Draper classes correspond to nearly equal increments of redness as measured by the color index.” [Pg 193] The preceding eight chapters review the arguments and the observations that have established the connection between the spectrum of a star and its temperature. From an examination of the data there given it becomes clear that what the Draper system classifies is essentially the degree of thermal ionization. A. Fowler, in fact, makes the illuminating distinctions of “arc” ( to ), “spark” ( to ), and “superspark” ( onwards) stars. The table that follows contains, in concise form, the chief features by which the type stars of each class are to be recognized, although it is again emphasized that these were not actually measured as criteria for the Draper classes. The lines characteristic of each class serve, however, to specify its degree of thermal ionization.