Transgressive Variability. 435 



the seeds of 0. biennis and 0. muricata almost exactly 

 alike as to shape and size in spite of the considerable 

 difference between the seeds of a single fruit. The same 

 is true of another well-known character, the relation be- 

 tween the length of the corolla and that of the filaments. 

 I measured this in ten flowers of 0. muricata and in 

 twenty of O. biennis and found the mean for the former 

 species to be 14.6 and 8.3; for the other 10.0 and 5.5 

 mm., in both therefore a proportion of 100 : 55. In 

 O. Lainarckiana however this proportion was 100 : 44. 



The length of the fruits is dependent to a large extent 

 on the conditions of life. If we examine plants which 

 have been grown under diverse conditions, we find differ- 

 ences in the lengths of the fruits and we get series of 

 figures which bear a superficial resemblance to the fore- 

 going ones but are due to different causes. See the table 

 which follows. I measured the fruits in the ripe, prac- 

 tically dry condition, using in each plant the lowest cap- 

 sule on the main stem; the plants were collected in wild 

 localities, but in various places. The differences have no 

 specific value and are manifestly due to conditions of 

 nutrition. An opposite result might have been obtained 

 under reversed conditions. 



Space does not permit me to deal in the same way wn'th 

 the variability of the new species which have arisen from 

 Lainarckiana. The great majority of the characters are 

 without doubt transgressively variable ; unbroken series 

 could be easily made with the leaves of O. snblincaris at 

 the one end and those of O. lata at the other; or with 

 the fruits of O. oblonga at the one end and those of 

 O. rubrinervis at the other. But whenever a sufficient 

 number of individuals are dealt with, curves derived 

 from this material are found not to be monocentric but 



