14 
BULLETIN 1164, U. S. DEPARTMENT OF AGRICULTURE. 
and the coefficients of variation for characters determined by 
measurement are given in Table 1. 
Comparisons of the several populations in respect to the coefficients 
of variation for 28 characters determined by measurement and the 
standard deviations for 11 characters determined by grading are 
summarized in Table 2, which gives the data (1) for all characters 
and (2) for characters in respect to which the coefficients of varia- 
tion or the standard deviations differed by an amount equal to three 
and one-half or more times the probable error of the difference. In 
comparisons involving the parental populations the small size of the 
latter and the resulting large probable errors of their constants 
doubtless explain the relatively small number of cases in which 
significant differences are shown. 
Table 2. — Relative variability of the parental and hybrid populations of Holdon and 
Pima cotton as expressed by the numbers of cheiracters in respect to vhich one or the other 
population gave the higher coefficient of variation of a measured character or standard 
deviation of a graded character. 
Number of charac- 
ters showing dif- 
ference between 
the coefficients 
Parental comparison. 
Comparison of the hybrid with the 
more variable parent, whether Hol- 
don or Pima. 
Comparison of 
F 2 with Fi- 
Holdon with Pima. 
Fi 
F 2 . 
the standard de- 
viations. 
Holdon 
more 
vari- 
able. 
Holdon 
less 
vari- 
able. 
No 
differ- 
ence. 
Fi 
more 
vari- 
able. 
Fi 
less 
vari- 
ble. 
No 
differ- 
ence 
F 2 
more 
vari- 
able. 
F 2 
less 
vari- 
able. 
dSeV k°;f 
ence - able. 
F 2 
less 
vari- 
able 
No 
differ- 
ence. 
Showing any dif- 
29 
8 
7 
3 
3 
10 
3 
27 
15 
1 
33 
25 
6 

1 38 



Showing a signifi- 
cant difference l . 
3, 
i Difference amounting to three and one-half or more times its probable error. 
Holdon was clearly much the more variable of the parental popula- 
tions, having given a coefficient of variation or a standard deviation 
higher than that of Pima for 29 of the 39 characters. Considering 
only characters in respect to which the constants differed significantly, 
Holdon was the more variable in 8 and Pima the more variable in 
3 characters. 
Comparing the hybrid F x with the parental population which was 
the more variable for the character in question, the coefficient of 
variation or standard deviation for 10 characters was higher and for 
27 characters it was lower in Y l than in the more variable parental 
population. F 1 gave a significantly higher constant than the more 
variable parental population for only 3 characters, while for 15 char- 
acters it gave a significantly lower constant. 
F 2 gave a higher coefficient of variation or standard deviation than 
the respective more variable parental population for 33 characters 
and a lower constant for 6 characters, but while F 2 was significantly 
more variable in 25 characters, it was significantly less variable in 
none. 
F 2 was significantly more variable than F x in all but 1 of the 38 
characters determined on both populations. 
