PRODUCTIVENESS OF SELF-FERTILIZED CORN 
13 
Table 6. — Relative -productiveness of crosses between self -fertilized lines of corn as 
differences in yield per 10 plants over or under the yield of the corresponding 
check plants in the number of perfect hills stated — Continued 
Pedigree No. 
Number 
of per- 
fect hills 
Yield (pounds) 
Field row 
Check 
Difference 
between 
check and 
cross 
1 
2 
3 
4 
5 
No. 71 
2-2-2-4-6-2 X 7-4-2-6-2-2 
48 
81 
9 
103 
44 
19 
20 
54 
18 
70 
30 
6.06 
6.08 
4.21 
5.97 
5.91 
6.94 
6.67 
6.76 
5.99 
7.23 
6.82 
— 1. 63±0.30 
No. 35.... 
2-2-1-3-2-1 X 5-1-2-3-2-2 
-1.64± .23 
No. 66 
2-2-2-4-2-1 X 7-4-2-6-2-2. 
-1.66± 
No. 4 
2-2-2-4-6-2 X 5-1-3-1-1-2 
— 1. 67db . 20 
No. 17 . 
2-2-2-4-4-1 X 5-1-2-3-2-2 
— 1.75d= .35 
No. 3. 
2-2-2-4-4-1 X 5-1-2-2-1-3 
-2. 38± . 60 
No. 64c 
7_4_1_7_4_1 x 10-3-2-1-2-1 
-2. 38± . 60 
No. 8 
5-1-3-3-1-2 X 2-2-1-3-2-1— 
— 2. 95± .30 
No. 54.. . 
10-3-2-1-3-3 X 5-1-2-2-1-3 
-5. 52± . 60 
No. 15 
10-3-2-1-3-3 X 2-2-2-4-4-1 
— 5. 67± .24 
No. 58—. 
10-3-2-1-2-1 X 2-2-2-4-2-1 
-6. 46± . 42 
The average yields of the checks are shown in column 4. The 
-average difference between the yield of each cross and its correspond- 
ing check, together with the probable error of the difference, is shown 
in column 5. Yields of crosses lower than the check are indicated by 
a minus sign ( — ). The data are in pounds of air-dry shelled corn 
per 10 plants. A perfect stand would have been about 7,000 plants 
per acre. The values shown multiplied by 12.5 give an approximate 
&cre difference in bushels with the same plant yields under perfect 
stand conditions. 
The data on the crosses grown in field rows Nos. 15, 54, and 58 
are included only for completeness. Their behavior was so much 
like that of the self -fertilized lines of their pistillate parents that 
experimental error is suggested as a possible explanation. In any 
event the behavior of these crosses clearly is outside the range of 
normal behavior of the crosses as a whole. Except as shown in 
Tables 6 and 7, therefore, these crosses will not be considered as 
having been in the experiment. 
The probable errors shown for the mean of n comparisons were 
obtained by dividing the generalized probable error of ±0.60 pound 
for a difference in any single comparison in the experiment by -yfn-1. 
The generalized probable error was determined from the data on 
the 62 crosses for which there were four or more replications. The 
difference between the yield of a cross and its check in each repli- 
cation and the mean of these differences were determined. The 
deviations of the differences in the individual replicates from their 
respective means then were obtained. The 470 deviations so ob- 
tained were used to compute the probable error of any single differ- 
ence. This is similar to the method described by Hayes {2), differing 
chiefly in that the actual deviations were used rather than the per- 
centage deviations. 3 
3 The distribution of the 470 deviations within classes limited by multiples of the probable error, 
together with the expected theoretical distribution, is— 
Number of deviations 
Multiples of probable error 
±E 
±2E 
±3E 
±4E 
±5E 
Observed 
236.0 
235. 
383.0 
386.7 
449. 
449.8 
465.0 
466.7 
470.0 
Expected 
469. 7 
