2 PRELIMINARY REMARKS. 
As a preliminary step towards the proposed classification, the family may be divided 
into two principal sections, distinguished by the characters of the anthers. Thus we have :— 
Ist, Haplantheree, where the anther-cells are simply adnate, without connective, or with a 
soft one scarcely longer than they. 2nd, Symphyantheree, where the cells, usually much 
shorter, are collaterally agglutinated upon the face of a large horny connective, generally 
bifid at its base, and furnished at the apex with a soft pointed appendage; these anthers 
conniving in a cone, their apical appendages bend inwards, closing together by their 
margins, so as to form a complete covering over the stigmata, at the same time that the 
ten anther-cells, thus brought into juxtaposition, become placed opposite to as many 
nectariferous glands belonging to the clavuncle, and thus become agglutinated to them: 
we shall have occasion to refer again to this singular disposition. 
The above plan evidently occurred to Prof. De Candolle in his arrangement of the 
family ; but he did not follow it as a means of classification. Dr. Miller only adopted it 
to a partial extent. But there are other equally important points of structure, which 
serve to form the limits of the proposed subdivisions; among these are the direction of 
the convolution of the segments of the corolla, the relative position of the stamens, the 
differences in their form, the shape and the divisions of the disk, and especially the 
‘nature of the fruit and the development of the seeds. This latter much neglected feature 
is here employed as far as possible, because it ought to form an important element in 
every system of classification. It includes the situation of the hilum on the seed, and 
the direction of the embryo in regard to it. De Candolle and Müller rarely give any 
indication on this point; and Endlicher, usually precise, is equally silent about it; but 
it is here employed to the full extent that my own observations allow. In those cases 
where no information is recorded, analogy has been the guide until more certain data 
are observed. 
In the second class of this family we find many points of extreme interest, in regard 
to structure and function, which have not received the attention they deserve, and which 
call for especial notice. Among these is the part here called the clavuncle, which is a 
fleshy cylindrical process, attached to the summit of the style, and which has generally 
been regarded as a part of the stigma; but as the two stigmata, placed upon its summit, 
are fully developed, this idea cannot be maintained: it must be regarded as a distinct 
indusiate process, adapted to a peculiar function. Upon the exterior surface of the 
clavuncle we find five or ten parallel longitudinal glands, exuding a nectariferous juice. 
The singular juxtaposition of the anther-cells has been above alluded to—leaving 
a hollow space in the centre, closed at the summit by the s/egiwm formed by the 
conniving apexes of the connectives, at the same time that the ten basal forks of the 
connectives, tending downwards, allow a free ingress from below into this hollow space. 
By the growth of the style the clavuncle is carried upwards, being guided into the 
cavity by the forks; but it cannot go further and break through the valves of the 
stegium, as it is stopped by the broad basal peltate indusium, which, acting as a buffer, 
prevents its progress beyond the precise point required, so that the nectariferous glands 
now stand opposite to, and in contact with the ten anther-cells. The glands begin to 
exude their juice at the same time that the anthers emit their grains of pollen; and 
