CHOEISIS OK DEDUPLICATION. 205 



pedicel below (Fig. 389, 390) , and normally alternate with them : 

 alternate with these is a pair of large petals, deeply saccate, or 

 spurred below ; alternate with these, a pair of smaller petals 

 with spoon-shaped tips which cohere at the apex (the corolla 

 therefore of two circles as in the related Poppy family) ; alternate 

 with these, two phalanges or united stamen-clusters, of three 

 stamens each ; alternate with these is nothing, for the second 

 set of stamens is wanting ; alternate with this vacancy is a pair 

 of carpels wholly combined into a compound 2-merous pistil. 

 The statement itself explains the morphology. The three sta- 

 mens of each phalanx stand in the place of a stamen, and are 

 the divisions of one. In Dicentra the members of the phalanx 

 are almost separate ; in Adlumia (Fig. 393) and Corydalis the 

 undivided filament reaches almost up to the anthers. The middle 

 anther of the phalanx is normal, or two-celled ; the lateral 

 anthers are one-celled, as if halved. 1 



1 Eichler adopts this interpretation (proposed in Gray, Gen. Illustr. 

 i. 118), and applies it to the crucial instance of Hypecoura. In the flower 

 of this Old World genus, there are four apparently simple and complete 

 stamens, one before each petal : the simplest interpretation would be that 

 which the facts appear to present, viz. that both dimerous circles of stamens 

 are complete and normal. But Eichler in view of the early development 

 and the double vascular bundles of the stamens before the inner petals, and 

 some occasional slight disjunction of their anther-cells considers that 

 the interior stamen-circle is wanting here, no less than in the other genera 

 of the order ; that what here takes its place before each inner petal is a 

 stamen composed of the adjacent lateral member of the phalanx, congeni- 

 tally severed from the group to which it belongs and soldered into one fila- 

 ment, bearing the two one-celled anthers so brought together as to imitate a 

 normal two-celled anther. The organogeny of the blossom is thought to 

 favor this hypothesis ; and it certainly favors the view here adopted of 

 the composition of the three-membered phalanx of the family generally. 

 If this interpretation of Hypecoum seems far-fetched, it is no more so than 

 its exact counterpart, through which DeCandolle, Lindley, and others explain 

 the case of the rest of the family. Starting with that genus as the simple 

 type, they conceive that the stamen opposed to each inner petal is each 

 severed into two, and that these half -stamens attached to the sides of the two 

 intact stamens, thus producing the phalanges by coalescence. 



A good empirical conception of the formation, from a single leaf, of three 

 stamens in Fumariacese, or two in Cruciferae, is afforded by the petals of 

 Hypecoum, as illustrated by Eichler. The outer petals are slightly three- 

 lobed from the apex; the inner are deeply so and narrower. The mem- 

 bers of the next circle in the family generally are just such three-lobed 

 bodies, the tip of each lobe transformed into an anther. There is an ap- 

 parent congruity in the production by the symmetrical middle lobe of a 

 symmetrical two-celled anther, and of a one-celled anther by each unsym- 

 metrical lateral lobe or stipule-like portion. A fuller development of these 

 sides of the leaf, and non-development of the middle portion (somewhat 

 after the analogy of Lathyrus Aphaca, Fig. 219), with anther-formation, 

 would convert the leaf into a pair of stamens 



