NOSTRIL-FORM. 65 



(21 per cent are so massed in one table, 17 per cent in the other). Therefore, 

 in accordance with hypothesis we must regard the lower grade narrow 

 slit as recessive. Similarly, heterozygous X low nostril (table 47) should 

 give, on our hypothesis, 50 per cent low nostril. If that is recessive we should 

 expect a massing of this 50 in the first two grades; if dominant a greater 

 scattering. The former alternative is realized. Again, in the heterozygous 

 X high nostril hybrid (table 50) the upper 50 per cent will be massed or 

 scattered according as high nostril is recessive or dominant. Allowing for 

 the 50 per cent heterozygotes in the progeny, the 50 per cent of high nostrils 

 are scattered through at least 8 grades of the possible 10. High nostril 

 is dominant. Finally, extracted high nostrils bred together produce off- 

 spring (table 52) with a great range of variability (through all grades), 

 while extracted low nostrils (unfortunately all too few) give progeny with 

 grades 1 and 2 (table 53; fig. B, h). Accepting, then, the general prin- 

 ciple of the greater variability of the dominant character, we have demon- 

 strated conclusively that high nostril, or rather the factor that determines 

 high nostril, is dominant. 



Comparing tables 45 to 54, we see that recessive parents are character- 

 ized by a low grade of nostril and they, of course, tend to produce offspring 

 with a low grade. Similarly, dominants have a high grade and tend to 

 produce offspring of the same sort, while heterozygovis parents are of inter- 

 mediate grade and their children have nostril grades that are, on the aver- 

 age, intermediate. Without regarding the gametic constitution, we might 

 conclude, with Castle, that offspring inherit the grade of their parents, 

 and consequently it would be possible to increase the grade, perhaps indefi- 

 nitely, by breeding from parents with the highest grade. Considering the 

 gametic constitution of the parents, it is obvious that such a conclusion is 

 premature. To get an answer to the question it is necessary to find if 

 there is, inside of any one table, among parents of the same gametic consti- 

 tution, any such relation between parental and filial grades. This can be 

 determined by calculating the correlation between the grades of parents 

 and progeny. Such calculation I have made for table 48 with the result: 

 index of correlation, r= 0.018 0.032, which is to be interpreted as indicat- 

 ing that no correlation exists; and in so far the hypothesis of Castle proves 

 not to apply in the cases of booting and doubt is thrown on the significance 

 of his conclusion. 



Finally, if we throw together the frequency distributions of all tables 

 into one table (table 55; compare fig. B) we shall find the totals instruc- 

 tive. Table 55 shows that, when all results are thrown together, including 

 hybrids of all sorts, grade 2 and grade 9 are the most frequent and grade 

 6 is the least frequent, the frequency gradually rising towards the extremes 

 of the series. The same result appears in the individual series that range 

 from grade 1 to grade 10. What is the meaning of this result? It seems 

 to me to bear but one interpretation, namely, that there are only two centers 



