21(5 THE ANDKCECIUM OF MENTZELTA. . 
US an occasional arrangement in Merdzelia, where there are/t"^ stamens 
of the first degree, ten of the second, '^niffteen of the third, ten of 
these kst being superposed in pairs to the primary stamens, and five 
alternate with them, the aiTougement cannot fail to strike one as being 
the /?! 
then^/( 
relation to the petals as the fifteen stamens of the third degree in 
Mentzelia have to the primary stamens. This analogy is so very close 
that it seems impossible to doubt that whatever explanation holds good 
as to the one case must necessarily do so as to the other. If, therefore, 
my interpretation of the anJrceciura in the Mentzeliece be correct, it 
will follow that in Rosa and Geum. the andrcecium consists of staminal 
groups, of which the petals, ordinarily so called, are the apices, just as 
the false petals in Bartonia are the apices of its staminal groups. In 
liitbus the evidence of staminal groups of which the -petals are the 
apices is still more apparent, where the staminal evolution manifestly 
extends itself from five centres or points of departure, the centres being 
the so-called petals. As to Fragaria, Spiraea, and the like, their sta- 
minal arrangement only differs from that in Geiim or Eosa in the num- 
ber of stamens developed after the first range of ten. 
It will be seen from the foregoing, that, if my reasoning be admitted, 
all the Eosaceous forms above described must be considered as, strictly 
speaking, apetalous ; the androecium consisting of five confluent com- 
pound stamens with petaloid apices, alternate with the sepals. The 
genus AlcJiemilla^ it appears to me, stands in a most interesting rela- 
tion to these forms. Here the flower is also apetalous, but, instead of 
a whorl of compound there is one of simple stamens alternate with the 
sepals. In AlcJiemillaj we have, I believe, so far from an aberrant 
form of staminal arrangement, as is generally supposed, in reality a 
type — complete, though of elementary simplicity — of the andrcecium of 
a great portion of the family. 
A second series of Eosaceous andr(£cia is to be found in the genera 
Aremonia^ Agrimonia, Sanguisorha, and Poierium, The development 
of the flower in these forms, as detailed by Payer, shows that they are 
characterized by having a single whorl of stamens superposed to tJie 
sepals. In Aremoiiia and Agrimonia there is a true corolla of petals 
alternate with the sepals, which in Sanguisorla and Poierium dis- 
appears. In Aremonia and Sangimorha the stamens are simple, while 
