450 BOTANICAL GAZETTE [DECEMBER 
As a root’s weight is negatively correlated with its percentage 
of sugar and positively correlated with its total sugar content, an 
absence of correlation between one pair of characters in tables 
XXVII-XXXIT would signify a lack of correlation between each 
Average percentage of sugar 1h 1910. 
. 
S % 0 ™ 
{0 ee ee Ne mm ee a 
8 6 6 ee ee ee 
) 
2s 
Boa) 9 
wo 2? 
o 
© 8 9 
> ws , d 
c 2 
oc 
> 
tits l2 
Fic. 35.—To accompany table XXXI 
of the other pairs. The coefficients of correlation not only confirm 
this assumption but also have a corroborative effect upon one 
another. 
Figs. 31-36 furnish a graphic illustration of the character of the 
Fic. 36.—To accompany table XXXII 
w 
ee 
& s Mverage weight of rools inounces 190. 
So 
= ~ L a 24 2 
SS 
: & 1/0 
2 8 . 
Ad 2 i i 
NO 
oe 
material used and the reliability 
of the results. The trends of 
the empirical means of the sep- 
arate classes represented by 
dots and connected by broken 
lines show a fairly close agree- 
ment with the straight line 
regressions. Moreover, there 
are no marked deviations of 
empirical means from the sade lines except where the classes 
are small. Hence the distributions are fairly regular and the 
biometrical constants are probably reliable. 
We are not justified in concluding from the foregoing results 
that no real differences in yield and quality occur, however, 4S 
real differences may be obscured by fluctuations. The most we 
