Equilibriuni Between Selection and Nutrition. 567 



be counted, and recorded. About 60 plants with 5, 6 

 and 7 florets were left over. Amongst these a further 

 selection was made of those whose branches were richest 

 in 5-7 rayed inflorescences. Immediately after the weed- 

 ing out had taken place these plants were deprived of all 

 inflorescences which were either in flower or over, in 

 order that all their seed might result from pure fertili- 

 zation. Of the twelve plants thus treated I chose the 

 four strongest and most fertile as seed-parents for next 

 year's crop: their terminal inflorescences had 5, 5, 6 and 

 7 ray-florets respectively, and their lateral branches bore 

 heads with few rays. In 1894 I obtained from their 

 seeds 256 flowering plants and determined the curve from 

 them in the usual way. 



The figures I obtained are summarized in the table 

 on page 568. (See also Fig. 119 C.) 



The third experiment was carried out with Bidcns 

 grandi flora (Fig. 119 B). In this species the inflor- 

 escences are usually five-rayed, but the number, here 

 also, is subject to variation and wuthin limits similar to 

 those in Coreopsis. 



In the flowers of Dicotyledons the number 5 is as a 

 general rule remarkably constant, and probably in a 

 great many cases hardly subject to any fluctuations. The 

 question naturally presents itself : why is this number 

 inconstant in this case? This problem has however not 

 yet been investigated; a solution of it would of course 

 be of fundamental importance to the student of varia- 

 bility. ^ 



I obtained my seeds in the winter of 1891/92 from 

 Messrs. Haage & Schmidt in Erfurt, sowed a square 



^The question Is whether a cyclic arrangement diminishes the 

 variability of the number of the parts involved and if so: why? 



