372 
De Vries on the Onfjhi of Species 
The second generation of offspring, produced in 1891, contained many plants. 
Three hundred of these plants, which were examined in August, bore 
Leaves with 3 leaflets 7189 
Leaves with 4 or 5 leaflets ..... 1177 
8366 
or an average of four abnormal leaves per plant through the whole series of 
individuals. The abnormalities were more evenly distributed than in the first 
generation, since 80 per cent., instead of fifty per cent., bore some abnormal leaves. 
An observation later in the season showed that some half-dozen plants had 
produced leaves with six leaflets ; and the best plant of all produced finally 36 per 
cent, of abnormal leaves, including six with six leaflets. 
Here, after selection during two generations, the variability has increased in 
one direction, as the ordinary theory would lead one to expect : but there is again 
no evidence that the mean character of the generation has " regressed " towards a 
type with man}^ leaflets. 
From this time onwards the conditions of culture and of selection were changed. 
The seeds produced in 1891 Avere sown in a greenhouse, and were transplanted 
from seed pans, after the appearance of the third leaf, each into a pot of well 
manured garden soil. Only those in which the third leaf possessed a supernumerary 
leaflet were preserved, and of these there were only 18 out of several hundreds of 
seedlings ; we have therefore no means of comparing the mean abnormality of this 
generation with that of previous generations ; all we can learn from this and from 
the subsequent observations is that under conditions of culture favourable to the 
production of supernumerary leaflets the percentage of such leaflets among the 
extreme offspring of stringently selected ancestry increased. The large amount of 
destruction which occurred in this and in subsequent generations clearly shows, 
liowever, that there was no regression of mean character in the direction of a new 
specific type during any part of the experiment. Thus in 1893 the seed of the 
18 plants selected in 1892 produced 8409 seedlings, of which only 938, or less than 
30 per cent., exhibited the abnormal third leaf for which their parents were 
selected. 
In subsequent years a still more stringent form of selection was adopted. The 
seed produced by each parent was sown separately, and the percentage of seedlings 
in which the third leaf was abnormal was noted in each case. Not only were all 
seedlings rejected in which the third leaf was not abnormal, but the seedlings 
preserved were taken only from those families which contained a percentage of 
abnormal individuals. 
By proceeding in this way a race of clover has been established in which the 
modal number of leaflets is approximately five, and deviations occur with fairly 
symmetrical frequency in either direction. Leaves with more than seven leaflets 
