PART 111. | THE MECHANISMS OF FLOWERS. 317 
Thus, while the flat layer of honey in Umbelliferse is only suitable 
_ for short-lipped insects, the honey of Composite is not only 
accessible to them but may be sucked by Lepidoptera and bees 
also. 
_ 3. The pollen-mechanism agrees essentially with that of 
_ Lobeliaceze. The anthers cohere to form a hollow cylinder, and 
 dehisce introrsely, filling the cylinder with pollen, before the flower 
' opens. The two stigmas lie at first closely applied together in 
| the lower part of the anther-tube, and as the style grows they 
- brush the pollen out of the tube by means of the hairs on their 
outer surface. So in the first period the pollen is exposed to 
' insect-visitors, and in the second the stigmas separate and expose 
their papillar surfaces. 
The arrangement of the hairs and of the stigmatic papille 
shows some diversity in Composites. The former sometimes, as in 
 Lobeliacex, sweep the pollen before them, forming a simple ring 
' round the style at the base of the stigmas (e.g. Centawrea, Cirsiwm), 
or aggregated in a tuft at the extremity of the stigmas (e.g. 
Achillea, Chrysanthemum); sometimes they are spread over more 
or less of the surface of the style, and the pollen remains en- 
tangled among them (e.g. Leontodon). The stigmatic papille 
sometimes wholly or partially cover the inner surfaces of the 
stigmas, and sometimes are restricted to broader or narrower 
» at the margin. This sweeping mechanism, which is 
_ apparently inherited from Lobeliacezw, is perfected in many 
| Composite by a peculiar irritability of the filaments, which 
_ contract when touched by an insect’s proboscis and draw the 
anther-tube downwards, squeezing the pollen out at the upper 
end exactly at ics proper time for it to be carried to other 
flowers. 
Crose-fertilisation i in case of insect-visits is thoroughly ensured 
for the separate florets in the same way as in Lobeliaces, and 
' it is rendered in many cases very probable, and in others in- 
_ evitable, for the inflorescence as a whole. For in all cases in 
_ which the development of the hermaphrodite flowers progresses 
so slowly from the margin of the capitulum towards its centre 
- that one or more rows of florets in the female stage are always 
surrounded by several rows in the male, the crossing of separate 
_capitula must at least be effected by all insects which alight at the 
margin (Bellis, Chrysanthemum, ete.); but in many Cynaree the | 
florets open centripetally in such quick succession that the 
_ capitulum is for a time male only, and afterwards for some time 
' 
| 
