50 Salisbury . — Variation in Er ant his hyemalis , 
the three stamens opposite the outer perianth segments (Fig. i). The 
three members of each whorl dehisce in rapid succession, followed after 
a shorter or longer interval by those of the next whorl within. The number 
of stamens, then, and their order of dehiscence, indicates that the androecium 
is quite commonly whorled, but a pseudo-spiral appearance is brought 
about by mutual pressure which results in a lateral displacement comparable 
to that seen in the floating rosettes of Callitriche vernalis or the leaves sur- 
rounding the inflorescence of Cyperus alter nifolius. In other cases, however, . 
the androecium is obviously spiral, frequently with a divergence of T 2 T or T 2 ^, 
the eleven or thirteen parastichies being clearly recognizable. As was 
pointed out by Irmisch (loc. cit.), the flower of Eranthis is then either cyclic or 
hemicyclic, or when, as happens but rarely, the perianth is spiral the flower 
may be completely acyclic. 
Eichler ( 1878 , p. 169) describes the androecium of Eranthis as most 
frequently consisting of rows of three stamens each, opposite the petals, 
alternating with which are six rows of four members, a total of forty-two 
stamens. Not only, however, have we failed to confirm this arrangement, 
but the number of stamens in the specimens examined seldom reached this 
figure. 
The gynaeceum consists of a varying number of carpels, usually either 
five or six. In the latter case two whorls of three members each are clearly 
recognizable. 
A drawing of a typical flower is shown in Fig. 1, the roman numerals 
indicating the order of dehiscence of the stamens. In Fig. 3, C, is shown by 
means of a diagram a typical cyclic flower. 
(b) Meristic variation T 
(1) The perianth. The ‘curve’ given in Fig. 4 shows the numerical 
variation exhibited by the perianth. As will be seen, the lowest number of 
members observed was five, but in such cases, with few exceptions, the place 
of the sixth member was taken in the outer whorl by a bract. The actual 
variation, apart from transformation, is then almost entirely in the direction 
of increase, the maximum number observed being nine. This unilateral 
character of the variation ‘ curve 5 for perianth members is frequent, and de 
Vries ( 1908 , p. 740) in Ranunculus bulbosiLs has isolated a concealed secondary 
summit to which this asymmetry is to be attributed. 
If a large number of flowers be examined one is struck by the repeated 
occurrence of perianth members showing all degrees of lobing, examples of 
which are shown in Fig. 5, E-G, and Fig. 6 , C. Nearly every condition, from 
a mere notch to two almost completely separated halves, has been observed. 
Most commonly the two lobes lie in the same tangential plane, but less 
frequently the lobes are large and overlap to a considerable extent. When 
