504 Willis . — Endemic' Genera of 
any really large islands received by water, the prediction woiild probably 
break down. 
If endemic genera are survivals, one cannot predict to what families 
they will belong, except that it would seem highly improbable that they 
should belong mainly to the large and what are usually called the 4 success- 
ful 5 families, like Compositae or Rubiaceae. It would seem more probable 
that they would show a tendency, at any rate, to belong to families that are 
small, or of broken distribution, such as we have been accustomed to con- 
sider as unsuccessful and more or less moribund. In any case, one would 
expect some marked differences in composition of the list from that of the 
mainland. 
If then, examining the endemic genera of islands, we find them 
to be a miscellaneous assortment, we may imagine either that the islands 
received their floras mainly by casual oversea transport, or that the endemics 
represent survivals, and if we also find that they show a distinct tendency 
to belong to the small and broken families, we may then with a fair degree 
of probability accept the second of these suppositions. 
If, on the other hand, endemic genera be young genera in the earlier 
stages of spreading, as is the case if we accept the hypothesis of Age and 
Area, we shall expect them (provided of course that the connexion was 
mainly by land) to appear in families in proportions not dissimilar to the 
proportions of genera appearing in those families at the present time. 
There should be among them many Compositae and Rubiaceae, few Mag- 
noliaceae or Myristicaceae, more Melastomaceae than Ranunculaceae, and 
so on. 
But we may go farther than this. As I have already pointed out 
(21, p. 34), if Age and Area holds, then on the whole a large family will be 
older than a small one of the same circle of affinity. And as the islands 
were mostly cut off from the mainland at a remote period, the families that 
reach them will on the whole be the older, that is to say, rather the larger 
than the smaller. And as the larger would be more likely to get to them 
first (as being on the whole older than the smaller), they will tend to have 
rather more genera in proportion upon the islands. One will therefore 
expect the proportion of endemic genera upon the islands to be if anything 
rather greater in the large than in the small families. 
Other factors will also come in to influence the result. Some families, 
like Compositae or Orchidaceae, may be well able to travel across wide 
stretches of water, and may thus be better represented than their age 
would indicate. If herbs are younger, as there is good reason to suppose 
( 5 ), we shall expect many herbaceous families, even though they may be 
large, to be badly represented, and so on. 
This second prediction, then, if it prove successful, will on the whole 
show that the bulk of the connexion to the islands at the time they 
