DORSIVENTRAL FLOWERS 131 



only it is quite chlorotic, and has no trace of chlorophyll. These white 

 leaves make the inflorescence conspicuous from afar. The inner flowers 

 have five linear small sepals.' Pogonopus Ottonis, another rubiaceous plant, 

 which I met with subsequently in Venezuela, exhibits similar features. 

 In these Rubiaceae then only one sepal in the peripheral flowers is 

 differently constructed, the flowers remain otherwise radial. When one 

 endeavours to give a causal explanation of the occurrence of dorsiven- 

 trality in flowers these cases must evidently be distinguished from the 

 others. In most of the examples of this kind the flowers are laid 

 down as radial structures, but by the preference given to the outer 

 part of the perianth they become dorsiventral. In the 'essentially' 

 dorsiventral flowers also the dorsiventrality comes about either through 

 the different construction of the parts of the flower which are laid down 

 radially, or through the inner parts of the flower being laid down in 

 different number or with a different construction from the outer ; often 

 enough both processes are combined, but then usually in such a way 

 that a dorsiventral flower results, or, in other words, the relationships 

 of symmetry of the several whorls of the flowers do not change inde- 

 pendently one of the other. Every systematic work supplies examples 

 of this, and especially Eichler's ' Bliitendiagramme.' I need only refer 

 to Fig. 82, copied from Eichler, which shows how a flower laid down 

 radially may by differences in construction of the androecium become 

 dorsiventral. The flowers of Commelina, like those of most monocotyle- 

 donous plants, consist of five trimerous whorls, a radial arrangement 

 which permits of symmetric division in three different planes ; but of 

 the six typical stamens only three are completely formed, the other 

 three, indicated in the diagram by crosses, are sterile, and diverge also 

 in the form of their cruciform four-lobed anthers from the three fertile 

 ones. The flower has therefore become dorsiventral and can only be 

 divided symmetrically in one plane, and associated with this we observe 

 that the fertile stamen through which the plane of symmetry falls differs 

 in construction from the other two by its possession of a broader con- 

 nective. The same change of symmetry would have been brought about 

 if the three outer stamens had been suppressed, as often happens. It 

 is however by reduction in the number of the carpels that the whole 

 symmetry of the flower is chiefly influenced, as will be pointed out 

 more particularly when the morphology of the flower is discussed, at 

 the same time most striking expressions of change of relationships 

 of symmetry are observed in the construction of the flower-envelope, 

 especially the corolla, in proof of which I need only refer to the 'labiate,' 

 ' ligulate,' ' personate,' and ' calcarate ' flowers. 



The relationships which I have just briefly sketched have from an 

 early period begotten explanations which have been partly teleological, 



K 2 



