Coluinelli.ice»e and Ciunuhitaccse. Ill 



meroiis, anatropal, the greater part of them e\teiiclin<!; liori- 

 zontally iVuin the placiiita, those on its uj)|)er part soiiieuliat 

 ascendinjr, those on its lower ratlier depiessed ; raphe in the 

 horizontal ovules apparently lateral. 



The stamens, however, of the Cucurbitacepe frequently consist 

 of three, one of which is only one-celled ; and the others beinj^ 

 two-celled, and the connective in some instances deeply forked, 

 has given rise to the opinion that the real number is five, and 

 that they are triadelphous, and consequently that the anthers 

 are only one-celled ; but that they are in such genera only three, 

 the one-celled anther of Bryonia dioicn places beyond doubt. 

 The smallest of its three stamens (PI. VI. fig. 22) has all the cha- 

 racters of being imperfect and bearing only half an anther, as it 

 is one-celled and always obviously unilateral, standing sideways 

 in the tube formed by the calyx and corolla, having the tortuous 

 anther-cell on one side, and showing the naked connective on 

 the other, which in other instances of half-anthers (as in Ma- 

 ranta) is not always the case, from the filament becoming more 

 or less twisted*. The stamens of the genera having this trian- 

 drous character while the petals are five, should rather be regarded 

 as having relation to the three-celled ovary which remains rudi- 

 mentary, and if so, such genera may be compared with Colu- 

 mellia, where, as the stamens do not appear to have any fixed 

 relation to the segments of the corolla, they may be supposed to 

 alternate with the two carpels, as they are both attached to the 

 corolla laterally, the carpels being anterior and posterior. 



Comparing the anthers of Culumellia therefore with the two- 

 celled anthers of the Cucurbitacese, the only difference will be 

 that in the latter the cells are not so elongated, so that the two 

 extremities, instead of being brought together, as in Culumellia 

 (PI. VI. fig. 20), point in different directions ; but even this is 

 not constant, as in Bryonia dioica, whether the anther is one- or 

 two-celled (fig. 22), the opposite ends of the cell are bi'ought 

 round towards each other, although not so nearly approximated 

 as in Culumellia. They correspond also in the lower extremity 

 of the cell being less contorted than the upper ; and in stamens 

 of Culumellia accidentally smaller, the cells are not more tortuous 

 than in Bryonia, if so much so (fig. 21). And supposing the 

 connective of the two cells of the anther of Cucurbitacea3 to be 

 consolidated (which does occur in some instances where they are 



* The one-celleu anther of the Cucurbitaceae seems to be owing to a 

 tendency to unilateral stamens, as in Bryonia dioica it is ahvavs one of two 

 situate on the jiosterior side of the flower which is one-celled, the remain- 

 ing stamen being anterior; the half-anther is alternate with the petals, 

 while the other two are opposite, which would be accounted for by sup- 

 |)osing their relation to be with the rudimentary ovary. 



