254 Non-Isolablc Races. 



in spite of the repeated selection.^ The curve ]\1 has a 

 more normal shape than the corresponding curve "1892'' 

 in h'ig. 51 ; which is ohviously due to the fact that the 

 former represents a homogeneous group whilst the latter 

 is a composite curve embracing all the groups of tliis 

 culture. Curve V is related to curve M as it would be if 

 we were dealing with ordinary fluctuating variability ; it 

 is simply shifted to one side.^ 



It seems obvious that the race could still be improved 

 by sowing the seeds of those seed-parents the apices of 

 whose curves are at 11, i. e., that these curves could be 

 shifted still further to the right. I have made some such 

 sowings since 1892, but only on a small scale and not 

 without interruption. They w^ere not intended as a con- 

 tinuation of the experiment. The number of petals per 

 flower increased slightly, but the type itself Avas not es- 

 sentially altered. I especially never saw a trace of any- 

 thing like the origin of a double flower. 



In order to find out wdiether there was any likelihood 

 that the type of my race would in the near future mani- 

 fest an improvement I made the following calculation. ' 

 The 295 plants of wdiich the culture of 1892 consisted, 

 arose from the seeds of 21 seed-parents. I selected tlie 

 ten best of these parental groups and plotted the curves 

 for all the ofl^spring of each seed-parent. The curves 

 proved to differ very little from one another. Tlieir 

 apices all fell over nine petals, with one exception, wliicli 

 was over ten. I should say that in making the calcula- 

 tion I have left the groups which contained less than 300 



^ Whereas as a result of the selection of active characters the 

 whole curve is shifted; see Vol. T. p. 73, Fig. 18, and the third part 

 of the first volume. 



^ .Sec Yn]. T. Fie:. tt6. on page 536. 



