170 Pearl and Surface. 



plants ending in a given quintile to have remained in or near that 

 quintile. In this case also there is, on the whole, a greater tendencj' 

 for the extreme plants to have maintained their same relative size. 

 However, this tendenc.v is much less marked than in the cases cited 

 above. 



On the whole, the relative size (quintile position) of a plant at 

 the beginning of the season is a much better criterion of its probable 

 relative size for the whole season than is its relative size at the end 

 of the season. 



The second step in this analysis has been to study the mean 

 quintile position of each group of plants in the successive growth 

 stages. As shown in Figures 10 — 12, the group of plants starting in 

 a given quintile show more or less regi'ession towards the mean of the 

 population with the advance of the season. Owing in part, at least, 

 to the rather small number of individuals in each group there are a 

 number of irregularities in some of these plotted lines. 



However the main conclusion is entirely clear. Extreme variants 

 at the beginning of the season tend strongly, on the whole, to remain 

 extreme variants during the whole season. At the same time such ex- 

 treme variants do tend somewhat to regress towards the general po- 

 pulation mean as growth continues. This second tendency is, however, 

 by no means so strong as the first. 



The third step in the analysis of these variation curves has been 

 a study of the average relative size (mean quintile position) of the in- 

 dividual plants and of the variability of individual plants with respect 

 to relative size. The conclusions drawn from a study of these data are: 



1. The observed difference in the manner of growth of individual 

 plants and of groups of plants cannot be explained as the effect of 

 external, environmental factors. 



2. These differences are rather to be looked upon as the effect 

 of internal factors. 



3. The distribution of the average relative size (mean quintile 

 position) of individual plants is such as to suggest the random distri- 

 bution of these factors among the plants. The same thing is brought 

 out by the distribution of the relative measurements of plants starting 

 or ending with a given relative size (quintile). 



4. The simplest method of explaining these facts is to regard the 

 differences in the manner of growth as due to independent Mendelian 

 factors which are distributed at random in any population of open fer- 



