Campanulatce. 291 
already described. In some cases (many Composite) the stigmatic 
surfaces ultimately curl completely over and so come in contact 
with any pollen that may remain on the then almost withered 
andrcecium ; in this way self-pollination is ensured in the event of 
failure in cross-pollination,—although every chance of success is 
provided first for the latter process. 
The mechanism is both simple and effective, and the “ pollen- 
presentation tendency ” is reflected in the largest known natural 
group of flowering plants. Special modifications of the simple 
mechanism we have described are expressed in families like 
Goodeniaceae and Candolleaceae; but their distribution is strictly 
localised and their species are few. 
Such is the determinant tendency of Engler’s Campanulatae as 
a whole, Cucurbitaceae excepted ; and this, we shall urge, affords 
the justification for regarding them as the descendants of a common 
stock as recent in descent as that of the Rubiales, but distinct 
therefrom. To this matter we shall return shortly, 
Second to the pollen-presentation tendency is one which we 
have met in the Rubiales and their ancestors, the tendency to 
aggregation of florets ; but in Campanulatae it plays a much more 
extensive part. The method by which the aggregation is tttained 
is, moreover, not the same in the two cohorts ; thus the heads of 
Dipsacaceae are the outcome of cymose branching, while those of 
Compositae are essentially racemose. The importance of this dis¬ 
tinction is seen in relation to the less advanced types of in¬ 
florescence in the respective groups; in other words, the dis¬ 
tinction expresses itself in the difference between the two stocks, 
campanal and rubialian. In the latter the inflorescence-unit is 
conceived as being characteristically a cymose umbel; in the former 
the flowers would be arranged typically in a simple raceme, as 
occurs in many Campanulaceae. The essence of cymose branching 
is the discouragement, so to express it, of single vertically elon¬ 
gated axes, and so the flowers readily tend to be brought to one 
horizontal level; in the racemose branching the reverse is the case. 
As a result we are furnished with a fundamental criterion between 
the campanal and rubialian stocks, and so the critical character of 
syngenesis of anthers in the former receives a measure of sub¬ 
stantiation. In any case the umbellate type of inflorescence 
familiar, e.g., in Cornel, Ivy, Elder, Valerian, etc.,—all belonging to 
families relatively near to their respective stocks (see fig., p. 230)— 
rarely occurs in the family which includes Harebell and Lobelia. 
