Campanulatce. 297 
affecting both anthers and filaments. The climax is reached in the 
tropical American genus Cyclanthera, with over 30 species, i.e. 
(about 5% of the family), in which the anthers are fused into a ring, 
divided into two chambers along its length, and seated upon a 
column composed of the united filaments. 
There is, moreover, a tendency to oligomery of the andrcecium ; 
about 10% of the family is concerned. In Gerrardanthus (4 spp.) the 
andrcecium consists of four stamens and a staminode ; in Gurania , 
Anguria (over 70 spp.), etc., only two stamens are present in each 
flower. This tendency, coupled with the curiously twisted anther- 
sacs, might be urged to support the affinity of Columelliaceae (chap¬ 
ter VI, p. 154) with Cucurbitaceaae; but in view of the specialized 
habit and polycarpellary ovary of the latter, and their close 
relationships with Parietales, it seems improbable that the shrubby, 
bicarpellary Columelliaceae have any near connection with them. 
The affinity of Columelliaceae with Gesneracere is much more 
apparent (see chapter VI). 
The ovary in Cucurbitaceae is invariably inferior, and is com¬ 
posed usually of a number of carpels exceeding two. The usual 
number is three, but the gynaecium may be isomerous with the 
corolla, or it may be reduced to a single carpel. The number of 
ovules present is variable, from a single one to an indefinite number. 
The ovary is primarily unilocular, 1 the ovules being borne upon 
parietal placentae; but in typical cases the latter project so far into 
the cavity of the ovary as to meet and form a central column, and 
the ovules are carried upon the recurved productions of the 
placentae beyond this. The ovary thus becomes secondarily septate ; 
the structure can be studied conveniently in thin slices from a 
cucumber. 2 
The prevailingly sympetalous corolla, the cohesion in the 
andrcecium, and the multilocular inferior ovary with indefinite ovule- 
number suggest, superficially at any rate, affinity with the Cam- 
panulatae—a suggestion actually adopted in Engler’s system. We 
woul4 urge, however, that this resemblance does not lie very deep. 
The sympetalous corolla alone can afford no guide to affinity. There 
is, it would seem, a fundamental difference between the staminal 
cohesions in Cucurbitaceae and the syngenesis of anthers in Cam- 
panulatae. The latter is referable to a definite biological purpose, 
> Payer, loc. tit., pp. 442-3, and plate 81, figs. 15-37. 
8 See also Lindley, The Vegetable Kingdom, Third Edition, fig. 
CCXV11I., p. 313, in which the true structure is displayed 
with remarkable clearnes 
