NATURAL CLASSIFICATION. 129 



ease the general symmetry is completely changed, and this 

 phenomenon is of very great importance. 



If the filaments do not cling to the coroUar integument of 

 the calyx, then the calyx cannot be united with the germen. 

 But if the filaments are fixed in the coroUar integument of the 

 calyx, then the calyx may either stand free, as in the Melas- 

 tomeae, or may be united with the germen, as in the Pomaceae. 



When similar parts are united at their base, they always 

 shew a disposition to be more expanded at their lower part ; 

 and we may safely presume that they are not united, when 

 we observe that they become narrower towards the base. 

 Hence the Myrsineae, although they have a very deeply di- 

 vided corolla, are not at all inclined to produce polypetalous 

 flowers. 



Lastly, The difference of numerical proportions may often 

 be explained both from the abortion and from the union of 

 parts. When we compare two related plants, one of which 

 has five, the other ten filaments, either the former has taken 

 its character from an abortion of filaments, or the latter from 

 the union of two flowers. Although the Grasses have com- 

 monly three anthers, we sometimes see Grasses of the Hex- 

 andria Class, as if they had sprung from the union of two 

 belonging to the Class Triandria ; an idea which is strikingly 

 confirmed by the structure of the EhrJiarta. In like man- 

 ner we often remark, that plants which have usually six fila- 

 ments, exhibit twelve of them, as happens in Lythrum. On 

 the other hand, Hexandria Plants may lose a part of their 

 filaments, and appear as if they belonged to Tetrandria, of 

 which the genus Convallaria is a striking instance. Those 

 of the Class Tetradynamia, are perhaps originally of the 

 Class Decandria, and the fouf petals are to be regarded as 

 altered filaments, as in Thlaspi Bursa we sometimes observe 

 this return to the original structure. 



I 



