22 APPENDIX. 



stamens or partial suppression of the fifths are therefore no longer 

 found to offer unerring limits of demarcation between those 

 families^ and I have pointed out the existence of other characters 

 that can be more safely relied upon for this purpose^ viz. the 

 aestivation of the corolla and structure of the seed : these^ taken 

 in conjunction with the usual ordinary distinctions^ afford a 

 more certain guide. Judged by these rules, Anthocercis will be 

 seen not to belong to Scrophulariacecey and the position assigned 



natural 



be suggested. 



One very remarkable feature is the peculiar sestivation of its 

 corolla, which I find to be a constant feature in every species : 

 the somewhat unequal segments of its border are rolled inwards, 

 with their margins overlapping one another respectively; in some 

 the dextralj in others the sinistral edge remains uppermost, and 

 the segments thus folded are drawn closely together into a long 

 conical bud, with the apices somewhat imbricately interlaced : 

 this very peculiar mode of aestivation will be best understood by 

 reference to the diagrams I have given {loc. cit. p. 170). 



It has always four fertile stamens arranged in pairs, of which 

 one pair is longer than the other, with a shorter sterile filament 

 or a mere rudiment of one, or else a vacant space in the interval 

 between the longer stamens : the filaments originate near the 

 base in the throat of the short constricted portion of the tube of 

 the corolla, where they are most frequently ciliated and much 

 geniculated at their origin, forming a fornix that conceals the 

 ovary ; they 4hen assume a more erect position around the style, 

 curving outward towards the summit, and are all slightly inclined, 

 together with the style, towards one side of the more expanded 

 portion of the tube. 



I have frequently alluded to the fact of the extrorse position 

 of the stamens among the DuhoisiecSy where it occurs constantly. 

 In Duboisia, Anthotroche^ and another genus to be proposed, 

 each anther consists of a single hippocrcpiform cell ; but in An- 

 thocercis, although the anther is equally reniform and extrorse, 

 it is formed of two divaricated cui-ving cells, closely united at 

 their apex ; this bursts externally by two hnes parallel with the 

 margin. This extrorse position of the anthers appears to be 

 otherwise quite unknown throughout the Solanal alliance, and 

 would lead us to suspect that the Duboisie^ really belonged else- 

 where, did not all the other characters unquestionably place them 

 here. This anomaly is probably explained by a circumstance that 

 in the course of this investigation fell under my observation : in 

 A. gracilis I found a' single flower with its corolla much dis- 

 torted, where two of the stamens were hippocrcpiform, 1 -celled, 

 and extrorse, as in AntJiotroche, and the other two were bilocular. 



