The following year, in my first report on Inheritance in Poultry, I laid great stress on the imperfection of dominance, and this phenomenon has become more striking and clear in the subsequent years, until in the present paper it is recognized as the key to the explanation of many apparently anomalous types of heredity. The first case in the present work in which imperfection of dominance is considered is that of the hybrids between I and oo comb. Here median comb is mated with no-median. Each somatic cell of the hybrid--at least in the comb region--has only half the full determiner for median comb. The determiner is weakened, and so the median comb is imperfectly developed, namely, at the anterior end of its proper territory. The weakening varies much in degree in the heterozygote. The median comb may be reduced to 70 per cent of its normal length or it may not develop at all. The second case of imperfection of dominance is that of polydactylism. Extra-toe mated to normal gives extra-toe in 73 per cent only of the offspring in the case of the Houdans. Any trace of 6 toes (on one or both feet) is found in only 12 per cent of the hybrid offspring from a 6-toed Silkie parent. Certainly dominance here is very like blending. The third case of imperfection of dominance is that of syndactylism. No syndactyls were noticed in F1. My first conclusion was that syndactylism is recessive; but later studies have shown that it is dominant and that all matings of two syndactyl parents yield about 56 per cent syndactyl offspring. Rumplessness gives an illustration of how dominance may be so weak as to be absent altogether; so that from F1 alone the erroneous conclusion is drawn that it is recessive; indeed, in one strain, only faint traces of the character made their appearance in successive generations. Finally, winglessness is a character which _appears_ not to be inherited at all. Nevertheless our experience with rumplessness leads us to suspect that winglessness also is an impotently dominant character. Looking at the matter frankly and without prejudice, the question must be answered: Has not the whole hypothesis of dominance become _reductio ad absurdum_? What visible criterion of dominance remains, where dominance fails completely? All the usual statistical landmarks of proportional appearance in successive generations being lost, can one properly speak of dominance and recessiveness at all? Amid the general ruin of criteria, however, one means of detecting dominance remains. That extracted character which in F2 or subsequent generations shows in homologous[15] matings in some families a wide range of variability is dominant, while that extracted character which constantly, in all homologous matings, shows no or very little variation is recessive.