INTRODUCTION. xxvii 
such flowers as are lateral; so that care must be taken 
not to confound the latter when accidentally solitary in 
bracteated species with the naturally central flower. 
Distinctions drawn from the shape of the tube of 
the calyx can in no instance be employed. All varie- 
ties of form may be found in canina and tomentosa. 
The shape of the sepals may sometimes be consi- 
dered, but very rarely their degree of division. In 
bracteata they are broad and short with a point. Ar- 
vensis has them less elongated than they ordinarily are. 
Their persistence. however must always be attended 
to. It distinguishes Woodsit in some measure from 
carolina; characterizes all the Pimpinellifolie; and 
affords the principal diagnosis of Villose and Canine. 
Their reflection contributes to divide damascena from 
centifolia, and alba from some others. Their elongation 
is a principal feature in rubrifolia ; and is frequently to 
be attended to in distinguishing single specimens of 
cinnamomea from certain states of tomentosa. 
Petals seldom offer any remarkable differences. In 
most species they are concave and spreading ; in invo- 
luta they are turned inwards at the edge; in carolina 
crumpled ; in Lawranceana pointed. | 
Stamens vary only in number, and in this respect 
they can rarely be employed, except among second- 
ary characters. In rubrifolia and the rest of its divi- 
sion they are very few, and in Bracteate exceedingly 
numerous. Usually they remain adhering to the ori- 
fice of the fruit till it is decayed; but in sempervirens, 
moschata, semperflorens and some few others they drop 
off nearly at the same time as the petals. The last 
species may be distinguished from indica by this among 
other things. They are commonly about four times as 
d. 2 
