( 403 ) 
Looking at the various figures we notice that the studied charac- 
ters generally give fairly symmetrical curves, disregarding smaller 
irregularities. Only in a few cases, as with Anethum graveolens for 
the number of umbel-rays of plants on the sandy soil, fig. VI, for 
the number of lobes of the leaves of the well-fed plants, fig. VIII, 
and besides for the number of lobes of the leaves of Scandiv Pecten- 
Veneris of the fertile soil, fig. IX, the curve is markedly oblique. 
Only for the number of branches of the third order of /beris amara 
from the sandy soil, fig. V, a semi-curve has been obtained. 
Examining in the various figures the position of the two curves 
with respect to each other, it appears that they partly coincide. 
This means that in the two corresponding cultures plants are found 
in which the organ under consideration is as large or occurs in 
equal number in the well-fed and in the badly-fed plants. But at 
the same time they show that in one culture individuals occur, in 
which a definite part is so strongly or feebly developed, as are not 
to be found in the other cultures. The figures further show that in 
all cases except of the number of akenes of JZa/va vulgaris, fig. NI, 
the curve of the plants on sandy soil has been shifted to the left 
with respect to that of the well-fed plants. 
The observations now enable us to determine how great the 
influence of the nutritive conditions is in the various cases and 
whether this difference in development between the two cultures is 
the same for various parts of the same plant. 
Examining the figs. I—V, relating to the characters of /beris 
amara; figs. VI—VII of Anethum graveolens and XI—XIII of 
Malra vulgaris it appears that, whereas with the two former plants 
the shifting of the curve is very different in the various cases, it is 
about the same for the three characters of Malea vulgaris and for 
all three of them relatively small. So the curves enable us to form 
an approximate idea of the influence of various nutritive conditions, 
but a clear insight is only obtained when the curves are defined by 
definite constants and these are mutually compared. In this way it 
is possible to determine what influence feeding has not only on the 
median value of the character, but also on its variability. In order 
to obtain these values, the median value J/ and the quartile Q were 
deduced from the observations. From these the coefficient of varia- 
bility Ww which is a measure of the variability and enables us to 
mutually compare the variability of different characters, was calcu- 
lated by the method introduced by VerscrarreLT *). Also for the 
1) Ep. VerscHarreLt, |. ¢. 
28% 
