496 MR. O. BENTHAM ON CLASSIFICATION AND 



at first sight might appear of sufficient importance to separate 

 ordinally a number of large or small groups, or even single spe- 

 cies, such as habit, venation of leaves, inflorescence, anther- 

 cells turned outwards or inwards, styles distinct or united, ovary 

 completely or imperfectly divided into cells, fruit baccate cap- 

 sular or nucamentaceous, colour and consistency of the seed- 

 testa, albumen hard, fleshy, or more or less farinaceous, distance 

 of the embryo from the hilum, &c. But upon further investiga- 

 tion these characters prove to be either exceptionally isolated, 

 separating groups in other respects evidently nearly related, 

 or if two or more exceptional characters may be observable in 

 one genus or in a group of several genera, they are either so 

 variously combined or so gradually evanescent in intermediate 

 genera as hitherto to render unavailing endeavours to define 

 satisfactorily subordinate groups. I have no distribution into 

 tribes to propose as an improvement upon those adopted by 

 Baker ; but at the same time it must be admitted that some of 

 them are purely artificial. Several of the baccate genera are 

 more naturally allied to corresponding capsular ones than to each 

 other ; and the baccate character itself is often very uncertain, 

 as, for instance, in Cordyline. The outward direction of the 

 anther-cells, so useful in the definition of Iridese as well as of all 

 the Nudiflorae except Alismace*e, fails entirely for the separa- 

 tion of MelanthacesB— the genera usually included in that tribe or 

 suborder showing every degree between the outward, the lateral, 

 and the inward direction. 



It is true that, as already remarked, the limitation of the whole 

 order may in the same way be sometimes stigmatized as uncertain 

 or artificial. Hewardia, for instance, is in many respects nearer 

 to Iridese than to LiliacesD ; JBlandfordia resembles some Amaryl- 

 lideas rather than any other Australian Liliacea) ; the main charac- 

 ter separating Liliaceae from Juncese scarcely holds good, if we 

 compare the small-flowered Asparageae with the Xerotidese ; but 

 these cases are comparatively few, and wherever the boundary 

 line is drawn, it must be more or less arbitrary or artificial. 



Pontederaceae differ but little from Liliaceze, chiefly in their 

 aquatic habit and a slight irregularity in the flowers ; they are, 



and 



The 



perianth petal-like as in the first, but the outer one calyx-like or 



very 



