Dec. 10,1917 Breeding Sweet Corn Resistant to the Corn Earworm 563 
existing among individuals, is —0.254 ±0.024 and the intraprogeny 
regression of damage on prolongation is 0.72. 1 
The apparent effect of prolongation on damage within the progeny is 
thus found to be only about 70 per cent of the effect indicated by the inter¬ 
progeny regression. 
The closer relation found to exist among the means of the progenies 
might come about through the ipterprogeny correlation of prolongation 
with other protective factors. The general absence of genetic correla¬ 
tions in maize characters would render this explanation improbable, but 
any explanation of the relation of prolongation to damage should also 
apply to the relation between prolongation and number of larvae. It is 
not clear how prolongation can directly affect the number of larvae, and 
the coherence of prolongation with other protective characters is the only 
explanation that suggests itself. Tor example, the progenies with the 
greatest prolongation might be later in maturing. If this were the case 
and larvae became less numerous as the season advanced, the closeness of 
the interprogeny correlation between prolongation and damage would 
appear to be greater than it really is. To approximate the true effect of 
prolongation on damage, an attempt must be made to eliminate, as far as 
possible, the effects of other correlated characters. To do this, resort 
may be had to “ partial correlations/’ 
In the present example the partial correlation between prolongation 
and damage with respect to days to silking will give, so far as the data 
permit, the degree of relationship between prolongation and damage 
1 There are many difficulties in the way of securing a satisfactory expression for the intraprogeny correla¬ 
tions of damage and prolongation. 
To combine the crude determinations of all the individuals into a single population is to confuse the 
inter- and intra-progeny correlations. To avoid this it seems better to calculate the intraprogeny correlation 
for each of the progenies. 
There is a further difficulty in the choice of method. The customary product moment method, which is 
perfectly applicable to the means of the progenies, can not properly be used with the individuals of a single 
progeny owing to the pronounced skewness of the distribution of damage. In a great many of the progenies 
approximately one-half of the individuals have zero damage. This division of the plants into two groups, 
those that were damaged and those that were not, would seem to indicate that the biserial correlation may 
properly be used. 
Differences between the mean prolongation of first and second ears prevent the combining of first and 
second ears in a single correlation table, but the independent calculation of the coefficient for first and 
second ears in the separate progenies provides an added check on the reliability of the method. 
The method followed has been to calculate the biserial correlation in each of the progenies for both first 
and second ears. In most of the progenies the division was made between zero damage and a damage of one 
or more. In a few progenies a more equal division was secured by making the division between palants 
with a damage of one or less and two or more. No correlation was calculated where the smallest class fell 
below 10 individuals. 
A weighted average of all the coefficients is taken as the best single expression of the intraprogeny correla¬ 
tion. 
The mean intraprogeny regression was calculated by the formula. Regression of damage on prolonga¬ 
tion = 
where Rdp =average intraprogeny biserial correlation and <rd and <r/>=the square root of the mean of the 
weighted squares of the standard deviations. 
23717°—17 - 3 
