J. A. Harris 
473 
The frequencies of positive and negative differences, both for all differences 
and significant differences, for the individuals are shown in Table XI. 
TABLE XI. 
Comparison 
All Differences 
Significant Diffekences 
Positive 
Negative 
Positive 
Negative 
B— A 
16 
12 
5 
5 
C-B 
8 
20 
4 
13 
C-A 
9 
19 
2 
6 
The comparison of the variabilities calculated from the ovaries for all 28 trees 
is made in Table XII. 
TABLE XII. 
Comparison 
Difference 
and 
Probable Error 
Diff. 
E diff. 
Percentage Increase 
or 
Decrease* 
B-A 
+ •2630+ -0410 
6-41 
+ 9-49 
C^B 
- -2235 + -0389 
5-75 
-7-37 
C-A 
+ -0395 ± -0387 
1-02 
+ 1-43 
From these tables the same conclusion is to be drawn as from that for the com- 
parison for the variability of the number of ovules per locule, namely that both the 
eliminated ovaries, A, and the matured fruits, C, show a less variability than the 
general population from which a portion of the ovaries have been eliminated. 
These results are very clear in the comparison for the totals of the three series ; 
they are not so evident on the small individual samples with their higher probable 
errors. 
Coefficients of Variation. 
In the preceding section it was shown that a selective elimination brings 
about a different mean number of ovules per locule or per fruit in the eliminated 
and the matured ovaries. The ovaries which develop to maturity have a 
higher average number of ovules than those which are eliminated. Possibly, 
then, our conclusions concerning variability may be open to question because of 
the fact that we are comparing the variabilities of samples with generally higher 
means with those with generally lower means. The absolute variability is gene- 
rally roughly proportional to the mean value of a character. 
* Increase or decrease from the eliminated to the uneliminated or from the less mature to the more 
mature. 
