694 MR JOHN MCLEAN THOMPSON ON 



Further, the maximum length of stamen-filament was attained immediately before 

 the anther dehisced, and all filaments in each flower examined reached a practically 

 uniform maximum length. Thus a stamen, which in an earlier stage of the flower's 

 development had been much shorter than its neighbours, at a later stage came to equal 

 them in length. That is to say, that anthers which in the undehisced condition had 

 been at various levels, due to inequalities of filament-length, were all, at dehiscence, 

 found to stand on the same level, and thus the zygomorphic condition of the androecium 

 which had dominated the earlier stages of development was replaced by a seeming 

 actinomorphy at a later stage (fig. 11). That this is not a true actinomorphy will be 

 apparent when it is remembered that although, at some point of time, all ten anthers 

 may display pollen at the same level, nevertheless they did not dehisce simultaneously. 



It may be mentioned at this point that the flowers were incompletely protandrous, 

 and that the majority of the anthers had dehisced before the stigmatic surface was 

 exposed. 



An examination, then, of expanded flowers alone provided the following facts. In 

 the early stages the perianth was practically actinomorphic or only slightly zygo- 

 morphic, but the andrcecium was very markedly zygomorphic. In the later stages the 

 actinomorphic or slightly zygomorphic condition of the perianth was maintained, but 

 the staminal zygomorphy was reduced as dehiscence advanced. 



The condition of some of the floral parts in the bud stages is now to be described. 

 With regard to the perianth, it will suflice for the present to state that, in the bud, 

 both calyx and corolla were found to be markedly zygomorphic, but that, as the flowers 

 expanded, this zygomorphy was entirely or greatly reduced, so that, in the opened 

 flowers, the perianth was entirely or almost actinomorphic. 



The condition of the staminal disc need not be discussed at present, but certain 

 stages in the development of the androecium will be described. 



At an early stage the androecium was found to be actinomorphic. The stamens 

 were of equal length, each consisting of an anther and a very short filament. Soon, 

 however, a slight inequality in filament-length was perceptible, and, as a consequence, 

 the stamens were difl'erentiated into two groups of five, five with longer and five with 

 shorter filaments. The anther size, however, remained uniform. The posterior stamen 

 was short, the anterior stamen long, and the androecium as a whole consisted of alter- 

 nately long and short stamens. Scarcely had this diff"erentiation of stamens been com- 

 pleted when a further modification was introduced, for it was found that the filaments 

 were elongating with increasing rapidity as one passed from the posterior stamen, 

 anteriorly. 



The maximum rate of filament-growth was realised in the anterior stamen, the 

 minimum in the posterior. 



Morphologically, this eventuated in the production of an antero-posteriorly zygo- 

 morphic androecium (fig. 4). 



Immediately previous to the initiation of this zygomorphy the length of the 



