+------+----+----------------------------------+--------------------------------+------+---------------+ |Serial| Pen| Mother. | Father. |Total | Offspring. | | No. | No.+-------+----+--------------+------+-----+----+--------------+------+grade.+-------+-------+ | | | No. |Gen.| Races. |Grade.| No. |Gen.| Races. |Grade.| |Grade 1|Grade 2| +------+----+-------+----+--------------+------+-----+----+--------------+------+------+-------+-------+ | 128 | 728| [A]912| F2 | Houd. × Legh.| 2 | 1298| F2 | Houd. × Legh.| 1 | 3 | 3 | 3 | | 129 | 827| 298| F2 | Pol. × Min. | 2 | 3852| F3 | Do. | 2 | 4 | 5 | 5 | +------+----+-------+----+--------------+------+-----+----+--------------+------+------+---------------+ [A] _Cf._ Serial No. 12_a._ At the outset, then, we find (table 55) that even pure races with high nostril (Polish, Houdans), when bred together, vary much in the height of nostril (in perfection of dominance) and, in 2 per cent of the offspring, even show the typical narrow nostril (fig. B, _a_). On the other hand, in the narrow-nostriled races I have never obtained any such variation. The most deviation that I have seen from grade 1 is found in my strain of Dark Brahma bantams that frequently give grade 2. The variability of the high nostril, the stability of the low nostril, is _prima facie_ evidence that the former is due to the presence of a particular factor and the latter to its absence. [Illustration: Fig. B.--Polygons of frequency of grades of "openness" of nostril in offspring of various parents. _a_, Both parents pure bred dominants; _b_, both parents extracted dominants; _c_, one parent heterozygous, the other a dominant; _d_, both parents heterozygous; _e_, dominant by recessive; _f_, heterozygous by recessive; _g_, heterozygous by extracted recessive; _h_, extracted recessives; _i_, heterozygous by dominant; _k_, both parents second generation hybrids.] Next, the heterozygotes of F1 (table 46), may be appealed to; but they will give no critical answer. For expectation, dominance being imperfect, is that the hybrids will be intermediate, and the result will be the same whichever extreme grade is taken as dominant. The empirical mode in the distribution of the offspring is at grade 2. This implies much greater imperfection of dominance on the hypothesis that grade 10 is dominant than on the hypothesis that grade 1 is dominant; but this very fact supports the former hypothesis, since imperfection of dominance is obviously a feature of the character with which we are dealing.