1904] SHULL—PLACE-CONSTANTS FOR ASTER 371 
characters are proportionately influenced by every variation in the 
factors which determine their size, number, or other quantitative 
relation, and neither is affected by any factor which does not affect 
the other. The organs do not modify each other, but both are 
allected by the same conditions. Only confusion results from the 
failure to appreciate the difference between immediate and mediate 
correlation, as may be seen in BURKILL’s (1902) discussion of the 
correlation in the parts of the flower of Ranunculus arvensis, when 
he says that “reduction in the number of petals does not act as a 
reflex on the number of sepals in anything like the way in which 
the reduction of sepals may be said to promote reduction of petals.” 
If as the values of any pair of mediately correlated organs or 
characters are increased or decreased the correlation between them 
4 changed, it must mean that one or other of them becomes propor- 
llonately less sensitive to the causes producing the change of values, 
and becomes more fixed or more variable in its quantitative relations. 
Such a change is well illustrated by an interesting diagram presented 
by Burk (1902), in which it is shown that sepals, petals, stamens, 
and carpels of Ranunculus arvensis vary together, 7. e., are closely 
Correlated, in flowers having the total number of parts less than 19, 
utin flowers having a higher total number of parts the sepals become 
fixed in number at 5, and the correlation between sepals and the 
Parts which continue to increase becomes zero. In flowers with 
more than 22 parts the mean number of petals likewise becomes 
.. at 5. In flowers of still higher numbers of parts the carpels 
@ tendency to respond with proportionately less increase a8 
‘Sata with the stamens. It is plain then that in this edger 
: ¥ conditions which promote the formation of flowers with a high 
imber of Parts will tend to decrease the degree of correlation and 
MCE versa, 
c alee important fact which must not be bimeec oe 
bean . Coefficient of correlation do not necessari f ee a 
hess, nge in correlation. PEARSON (1903) has pointed ot F 
seneity M a population tends to increase the coefficient 0 
: tion, but of course such heterogeneity does not increase the 
degree of Correlation. It is probable that most of the marked 
“S which have thus far been observed in coefficients of correla- 
ee a ee 
