308 ILLINOIS ACADEMY OF SCIENCE 
antennal formula have means even below the average. On 
the other hand the offspring of the 1.80-1.89 class have the 
highest progeny mean of any parent class. 
If we compute the average fraternal mean for all the off- 
spring of the eighteen parents considered in this isolation, we 
find it to be 1.823 which is only 0.023 above the mean obtained 
for the line, showing regression to be practically complete. In a 
previous paper (Ewing, 1914 b), the writer found by com- 
puting the regression according to parent-class method that 
it was more than complete. The correlation table substan- 
tiates my earlier conclusions in regard to the negative results 
of selection in this subline.. 
An Analysis of the Results of Selection Made in an Attempt 
to Increase the Length of the Cornicles in Aphis avenae. 
In isolation 3, selections were made in an attempt to increase 
the absolute (not relative) length of the cornicles in Aphis 
avenae. Selections were begun in the 15th generation and were 
made in each generation up to the 26th. Eleven selections were 
made in all, and sixty-seven progeny obtained. In Table II 
there is given a parent-progeny correlation table for the cor- 
nicle length for these eleven selected generations as well as fer 
several generations previous during which no selections were 
made. In these latter generations work was done on the de- 
termination of the cornicle length of the genotype. 
This table shows a strong correlation between parent and 
immediate progeny in regard to cornicle length, and the graph 
which was made in my previous paper (Ewing, 1916, p. 59), 
showing the results of selection on cornicle length, for the 
most part appeared to show the same thing, In fact, I was so 
certain of positive effects after six or seven selections that I 
presented a seminar paper on my success in creating a new 
strain of plant lice through selection in a pure line. Contin- 
uing the work a few generations farther, however, it was found 
that there was almost a complete regression in regard to 
cornicle length, and this in the face of rigid selection. Con- 
sidering the fact that in these latter generations there was 
such a strong regression, I am inclined to attribute the earlier 
gains in cornicle length not to selection, but to environmental 
