56 
Salisbury . — - Variation in Eranthis hy emalis, 
garded, the number of stamens is again a multiple of three. Thus in three 
separate flowers, in each of which one forked stamen was present, the total 
numbers (inclusive of the forked stamens) were 30, 33, and 36. 
Having regard to the method of development of stamens and that the 
anther is the first part formed, it is evident that these bifurcated examples 
must have originated as separate papillae, though probably derived by the 
fission of a single primordiurn, the two halves of which subsequently fused. 
Here too, then, as in the perianth, we appear to have the two antagonistic 
tendencies of multiplication and maintenance of the trimerous condition. 
The large proportion of individual flowers with 24, 27, and 30 stamens is 
the outcome of the one, whilst the occurrence of numbers which are not 
a multiple of three is probably often the outcome of the other. Besides the 
examples, figured stamens have been found which were branched from the 
extreme base and consequently only a stage 
removed from two separate structures. 
Eranthis hyemalis is by no means alone 
amongst the Helleboreae in exhibiting a trimer- 
ous tendency, for, apart from the examples 
given later, Church (loc. cit.) states that the 
gynaeceum of Helleborus foetidus usually con- 
sists of three carpels, whilst the stamens vary 
in number from 30 to 54, with 45 as the normal 
condition. Both extremes and mean, be it 
noted, multiples of three. 
(4) The gynaeceum. The carpels in the 
specimens examined varied from 3 to io, though 
Church (loc. cit.) gives the range as from 3 to 
11, and here again importance is probably to 
FlG * 8 - Fused carpel and b e attached to the fact that the minimum 
stamen of Eranthis hyemalis (a) 
and branched carpel of Helleborus number is three. Although six carpels very 
niger (b). frequently occur, rather more individuals were 
found to possess five carpels, though Church 
(loc. cit.) gives six as the prevailing number. The gynaeceum would thus 
appear to be more specialized in the direction of reduction than the androe- 
cium, a condition that is by no means uncommon in the group. Where six 
carpels are present, they occur in two whorls of three members each, but 
where the number is five they appear to be grouped in a single whorl. No 
case of branching or fusion of carpels has been observed here, but in the 
closely allied genus Helleborus joined carpels do occur (cf. Fig. 8, B, showing 
two of the carpels in H. niger joined together), which suggests the possibility 
that departures from the strictly trimerous condition may be the result of 
congenital fusion of carpel-rudiments. The number of ovules in each carpel 
ranges from six to ten, but is most commonly eight or seven (cf. Table IV)- 
