BARBERRIES. 787 



throughout the genus, we may assume the ancestral herbaceous 

 barberry to have had the same peculiarity, but how this remark- 

 able degree of sensitiveness could have arisen is not so clear. It 

 would seem as if insect agency in some way or other must have 

 brought about a movement having such an obviously purposeful 

 relation to insect visits ; but when we reflect upon the almost uni- 

 versal absence of a similar movement in flowers similarly visited, 

 and the very questionable usefulness of the pronounced irri- 

 tability of " sensitive " leaves, it is apparent that such a simple 

 general explanation really explains very little. The few con- 

 jectures that the writer has to offer on the subject will be best 

 understood after we have considered what may probably have 

 been the evolution of certain structural peculiarities of the 

 flower. 



The anthers (Fig. 3, A), opening as they do by little valves 

 hinged at the top, present a form of dehiscence confined entirely 

 to the Berberidacece,, the Lauracem, and a few other nearly re- 

 lated families not represented in our native flora. Within the 

 Berberidacece, all the genera except Podophyllum (the May apples) 

 and Nandina have the anthers thus characterized ; hence, it is 

 clear that the stamens of the ancestral berberis were already 

 of this peculiar type, and so the antecedent stages should be 

 thought of as occurring in that line of berberidaceous herbs 

 which were the forerunners of the barberries. The herbaceous 

 genus Podophyllum contains species exhibiting a difference in 

 the stamens which affords us an important clew for the under- 

 standing of what these antecedent stages may have been like. 

 In P. Emodi (Fig. 13, A) the dehiscence of the anthers is by a 

 longitudinal slit down the middle of each lobe. In P. peltatum 

 (B), our common species, there is likewise a vertical slit, but it is 

 so near the inner side of the connective that there appears to be 

 but one valve to each lobe, the other inner valve having been 

 reduced to a mere vestige. To connect this condition with that 

 common elsewhere in the family, we need only suppose the at- 

 tachment of the enlarged valve to become gradually narrowed by 

 a continuation of the slit from below (C) until there remains only 

 the small hinge we find in the barberry stamens of to-day (D). 

 It deserves notice that the hinge, instead of being quite at the top, 

 is nearer the back of the anther, which is what might be expected 

 according to the hypothesis. 



In the other families we have mentioned as exhibiting a val- 

 vular dehiscence of the anthers there is found almost invaria- 

 bly a pair of nectar glands on each filament (see Fig. 15). Now, 

 it is a curious fact that in certain of the less highly developed 

 mahonias the filaments are each provided with a pair of append- 

 ages (Fig. 14, A) similarly placed but being, so far as we know, 



