64 Salisbury . — Variation in Er an this hyemalis , 
multiples of three, the only notable exception being the number twenty- 
seven, of which there were only eleven examples ; but this is doubtless corre- 
lated with the large number of flowers exhibiting twenty-eight stamens 
(viz. twenty-one), many of which are probably the outcome of complete 
fission of one of the staminal rudiments. Hence we get an increase of the 
higher category at the expense of the lower. 
We thus have again a ‘ curve ’ which may be regarded as consisting of 
a number of secondary variation * curves each having some multiple of three 
as its mode. Departure from the mode is to be interpreted as an outcome 
Fig, 12. Ficaria verna , variation ‘curve’ for androecium. 
of fission or of fusion. Here, as in E rant his , branched stamens are not in- 
frequent, and examples are figured in which two or even three anthers occur 
on the same filament (Fig. n, A and B). Rarely small abortive stamens 
have been observed, so that complete suppression may sometimes have 
taken place. 
(4) The gynaeceum . Here again, although the feature is less marked 
(cf. Fig. 13), the maxima on the variation ‘curve’ are chiefly situated at 
multiples of three, the most frequent number being 1 5. The general trend 
exhibits the same markedly asymmetrical form so frequently characterizing 
that of the perianth. 
(b) Correlation. 
Tables V and VI bring out the fact that there is a marked correlation 
between the different whorls as regards either increase or decrease, so that 
we cannot attribute an augmented corolla to staminal transformation, 
