12 THE COLOURS OF FLOWERS. 



we find the filaments growing flatter and broader, and 

 the pollen-sacs less and less perfect ; next we find a 

 few stamens which look exactly like petals, only that 

 they have two abortive anthers stuck awkwardly on to 

 their summits ; and, finally, we find true petals, broad 

 and flat, and without any trace of the anthers at all. 

 Here in this very ancient though largely modified 

 flower we have stereotyped for us, as it were, the 

 mode in which stamens first developed into petals, 

 under stress of insect selection. 1 



F;c. 3. Transition fr:m stamens into petals in white waterlily (Nymphara alba): 



" But how do we know," it may be asked, " that the 

 transition was not in the opposite direction ? How 

 do we know that the waterlily had not petals alone to 

 start with, and that these did not afterwards develop, 

 as the Wolfian hypothesis would have us believe, into 

 stamens ? For a very simple reason. The theory 

 of Wolf and Goethe is quite incompatible with the 

 doctrine of development, at least if accepted as a 



1 The waterlilies belong to a very ancient type, in some respects 

 partially intermediate between Monocotyledons and Dicotyledons ; but 

 the comparative unification of their pistil shows them to have under- 

 gone considerable modification. 



