therefore, appear more distinctly when individuals are grouped together which have the same breadth of head. I have grouped the material in four classes, with the result that the double maximum of frequency, corresponding to the breadth of face of the parental types, appears more strongly marked in every class. Therefore we must draw the important inference that the face of the offspring has a tendency to reproduce one of the ancestral types not an intermediate type. The effect of intermixture in this case differs, therefore, fundamentally from the effect observed in the measurements of stature.
When comparing the average breadth of face for Indians, half bloods, and whites, another interesting phenomenon may be seen. The average breadth of face of the half blood stands between that of the Indian and that of the white, but nearer the former. When computing this average from year to year, it is found that the same relation prevails throughout from the fourth year to the adult stage, and in men as well as in women (Fig. 7).
The relation of the three groups remains unchanged throughout life. The amount of white and Indian blood in the mixed race is very nearly the same. We find, therefore, the remarkable fact that the Indian type has a stronger influence upon the offspring than the white type. The same fact is expressed in the great frequency of dark hair and of dark eyes among half bloods.
Two reasons may be assigned for this fact. It may be that the dark hair and the wide face are more primitive characteristics of man than the narrow face and light eyes of the whites. Then, it might be said that the characteristics of the Indian are inherited with greater strength because they are older. It must, however, also be considered that half bloods are almost always descendants of Indian mothers and of white fathers, and this may have had an influence upon the race, although there is no proof that children resemble their mothers more than they resemble their fathers.
In carrying out the comparison of breadths of face it would be better to study the curves of distribution for each year, but the