The Genetic Coefficients of Specific Difference

For the precise study of evolution of populations, races, or species, nearly every problem sooner or later requires some measurement of the morphological divergencies in the groups under observation. This is equally true and the problem is fundamentally the same whether one be studying very closely related species of Drosophila (Dobzhansky and Mather, '39), varieties of gall wasps (Kinsey, unpublished), fields of irises (Anderson, '36a), or the races of man (Pearson, '26, and various other authors). It is usually taken for granted in such studies that any measurable feature or features of the organism will serve equally well as a measure of likeness if only the records be made with care and treated with the precise methods of biometry. Improvements have recently been made by considering differences in groups of measurements, the data being combined crudely (Anderson, '36a, '36b, Anderson and Hubricht, '38) or by refined biometrical techniques (Fisher, '36b). These methods are all based on the tacit assumption that species differences are expressed more or less at random. A study of such differences has convinced us that their morphological nature renders these methods relatively inefficient. Species do not differ in a random manner. They differ in a peculiar and subtle way. If any two closely related species of the flowering plants are examined critically it will be found that they differ as a whole by two sets of harmonically integrated tendencies (Anderson and Whitaker, '34). Such a conclusion, however, is of little use in quantitative work. In section I, therefore, there is developed a precise mathematical