SELECTION IN PLANT BREEDING 199 
by this method of selection even with a crop that is often cross-fertilized. 
But the diagrams show other facts. The published records show that 
the variability of the race was but little, if any, reduced by continuous 
selection. With extreme variants comparatively as far removed from 
each year’s type, available for planting in each successive generation, 
the gain each year should have been at the same rate, if the Darwinian 
interpretation of the role of selection were correct. On the contrary, 
we notice that the regular curve fitted to the crop averages for ten gen- 
erations, is first concave showing great progress made by selection, is 
later convex as progress becomes slower, and last becomes horizontal 

Fic. 4. DIAGRAMMATIC REPRESENTATION OF THE RESULTS OF THN ILLINOIS AGRI- 
CULTURAL EXPERIMENT STATION in selecting for high and for low oil content. Y, per 
cent. oil in crop; X, generations; h, high oil strain; 1, low oil strain. 
as no more progress results. It is very evident that the original stock 
was a mixed race containing sub-races of various composition inter- 
mingled by hybridization. Selection rapidly isolated these sub-races. 
The isolation was practically complete at the eighth generation in the 
case of the protein strains and the ninth generation in the oil strains. 
After this selection accomplished nothing. That the effect of selection 
was simply the isolation of a sub-race and not a continuous response, 
is further demonstrated by the fact that in 1903 another plot was 
started with seed from the isolated high oil strain. After four years’ 
cessation of selection, the average composition of the crop remained the 
same, showing that after complete isolation of a homogeneous type no 
retrogression of the selected character occurs unless intererossing with 
mediocre strains takes place. Fluctuation in composition still appears, 
but this is the non-inherited kind produced by external conditions. 
