: S PHA NER OGAMS. 



stamen by two or more is termed by Payer Dedoublement^, by Eichler and others Colla- 

 teral Chorisis, and must apparently be considered as a branching of very early origin. 

 This view is confirmed in this case by the fact that in the Grucifer- Atelanthera, the 

 median stamens are only split and the two halves of each provided with half-anthers, 

 'while in Grambe each of the four inner stamens puts out a lateral sterile branch, which 

 may be explained as the commencement of a further multiplication of the stamens 

 such as actually occurs in the Grucifer Megacarpaea and in many Gleomeae. Even if 

 the way in which increase of the typical dimerous number of the inner whorl of stamens 

 has been brought .about be still obscure, it appears certain that the inconstancy of 

 the number of the members of the staminal whorl proves that in Gruciferae and Gle- 

 omeae a deviation has arisen in this part of the flower from the typical dimerous number, 

 \\ hile the other whorls have remained unchanged. The only deviation which occurs in 

 the gynaeceum of the Grucifers is in the genera Tetrapoma and Holargidium, where, 

 besides the two lateral carpels, two median ones are also produced, thus forming a 

 four-lobed ovary ^. 



An essentially different kind of increase in the typical number of the members of a 

 floral whorl may be caused by the formation in the still very young bud of new 

 members of the same kind between those already in existence and on the same zone 

 of the receptacle ; i. e. by what we have already described as the Interposition of new 

 members. This I found to occur, for example, in Dictamnus Fraxinella (Fig. 357), and 

 is represented in the diagram. Fig. 383, by the stamens of later origin being shaded not so 



Fig. 383.— Diagram of the flower c\{ Dictaininis Fraxinella (cf. Fig. 357, p. 493). 



dark as those of earlier origin. It may, I think, be inferred from Payer's descriptions 

 and drawings that the same process occurs in the nearly related genus Ruta, and in the 

 families Oxalideae, Zygophyllaceae, and Geraniaceae included in the same circle of affinity ; 

 'vi%. that in these cases also five stamens are interposed between those already in 

 existence. If the five interposed stamens are supposed to be removed, there remains 

 in these families a regular pentamerous flower with four alternating whorls each 

 consisting of five members, such as is found in the nearly related Linaceae and Bal- 

 samineae^. 



^ [The theory of an original dimerous symmetry in the flowers of Cruciferge has been pushed 

 still further by Meschaeff (Bull. Soc. Imp. Nat. Mosc), who regards the four petals as also the result 

 of a lateral dedoublement of a single pair (see J^entham, Ann. Address Linn. Soc. 1873). — Ed.] 



^ [Holargidium is a section of Draba. According to Bentham and Hooker the four carpels of 

 Tetrapoma are an abnormality not constant under cultivation. The same authors also mention the 

 occasional occurrence of a similar abnormality in Brassica and Nasturtium. — Ed.] 



^ Doll (Flora von Baden, vol. III. pp. 11 75, 1177) and others suppose that a whorl has 

 become abortive between the corolla and ovary in Rutaceoe and Oxalidere, a hypothesis which is 

 not supported by the history of development, and which is superfluous on our hypothesis. To 

 assume abortion merely because certain whorls do not alternate seems to me to be going too far. 

 Besides, the ten stamens of Epacridese and Rhodoracece cannot belong to two but only to one whorl 

 in wliich five are of earlier origin, and five have been interposed. (Compare Payer, Organogenic de 

 la fleur, pi. 118). 



