Experiments with Poultry 127 
independently of each other, there being no correlation between 
the characters usually associated with a particular breed, so far 
as their inheritance is concerned. 
One of the most curious results in crossing black and white 
fowls is the occasional appearance of a blue (or Andalusian) 
color, which consists of a minute patchwork of black and white. 
It has been known for some time that this color does not breed 
true, but a pair of such blue fowls gives rise to 25 per cent black, 
50 per cent blue, and 25 per cent white. The explanation of 
this is that the germ-cells of the blue individuals are black and 
white, hence the result; but the special condition that leads to 
the occasional formation of blue by the combination of black 
and white is not known. Possibly the presence of latent colors 
determines the result. 
The most recent experiments with poultry are those of Daven- 
port. His work confirms many of the results already obtained 
by Bateson and Hurst, and also establishes the relation of domi- 
nance and recessiveness for some new characters. It is these 
latter points that will be especially considered here. 
A cross was made between the single comb, black Minorca and 
white-crested, black Polish. These races and their hybrids 
are shown in Fig. 11; 1-6. In the hybrid (5 and 6) the comb 
is single anteriorly and bifurcated behind. There is much variety 
in the extent to which the comb is split. In fact, it was single in 
one case. Neither parent type can be said to dominate, the 
Minorca having a large single comb, and the Polish a much- 
reduced bifid comb. Davenport suggests that these two types 
of comb may both be dominant types that combine to form the 
Y-shaped comb. When the hybrids were inbred, there resulted, 
in a total of ror offspring, 29.7 per cent showing single comb, 
46.5 per cent Y-shaped comb, and 23.8 nocomb (oronly papillz). 
Two interpretations are possible. On the assumption of two 
dominant and two recessive types, viz. (1 a) median comb and 
(1 b) no median, and (2 a) no splitting and (2 b) splitting, the re- 
sults agree more or less with the expectation. But on another 
assumption the results conform even more closely, viz. if we 
