314 OBSERVATIONS ON THE NATURAL FAMILY
137] which points to this extremity cannot in strict propriety be described as directed towards the umbilicus. M. Cassini has not noticed the direction of the radicle ; either from supposing it constantly connected with that of the ovulum, or, which is more probable, from not having ascertained it. These distinctive characters may be considered as fully sufficient to authorise the separation of Boojndea from Composite ; yet the same differences exist between certain genera referred and reallv belonging to TLubiacece and the principal part of that order.
There are, however, three other characters unnoticed by M. Cassini, which distinguish the flowers of Boojrklece from the hermaphrodite flowers of the whole of Compositae ; namely, the accretion of the base of the style with the tube of the corolla ; the absence of the epigynous disk or nec- tarium ; and the longitudinal subdivision of each cell of the anthera by a " receptaculum pollinis," as in most other families, and of which, indeed, there seems to be the rudi- ment in the syngenesious genus Petrobium, described in the preceding paper.
In the partial cohesion of the antherae, in which they re- semble Jasione, they certainly differ from all known Com- positae : but as in certain Composite the antherae are very slightly connected or entirely distinct ; — this, though a re- markable circumstance, can hardly be employed as a dis- tinguishing character.
The principal characters in which Boopidece differ from the greater part, though not from the whole of Compositae, are the corolla being continuous, or not jointed, with the ovarium ; the anthera? having no membranaceous appendix at top ; and the undivided stigma.
Boojjtdece differ from Dipsacece in the vascular structure and valvular aestivation of corolla ; in the aestivation, inser- 138] tion, and connexion of antherae; in the absence of the partial involucrum ; and in having alternate leaves.
In adopting M. Decandolle's description of Dipmceai} they would differ also in the important character of " ova-
1 Elor. Franc. o me ed. vol. iv. ]>. 221.
�� �
