332 
Journal oj Agricultural Research 
Vol. XXX, No. 4 
experiments than in others shows that 
the conditions of those experiments 
were more favorable than in others for 
damping-off. In order to make the 
results of the different experiments 
directly comparable the data were re¬ 
worked. The mortality percentage for 
each fungus in each experiment was cor¬ 
rected, in much the same way that 
agronomists correct plot yields for field 
heterogeneity. 3 For example, if the 
average mortality per cent for the three 
standard lines in experiment B was less 
than the average of the same lines for all 
experiments, the mortality percentage 
for each fungus was proportionally 
raised. If the average mortality for 
them was more than it was for all 
experiments each mortality per cent 
was lowered. The corrected per¬ 
centages were averaged and constitute 
the last columns in the tables. The 
averages in the next to the last column 
of the tables are open to suspicion 
because the percentages on which they 
were based are uncorrected for the 
varying conditions of the experiments. 
The corrected averages in the last 
column are under suspicion because 
of (1) the small number of standard 
lines on which corrections are based; 
(2) the probability that the respective 
fungi are differently affected by changes 
in conditions, and (3) the somewhat 
questionable character of the assump¬ 
tion on which the method of correction 
is based. It is thus probable that the 
corrected percentages are corrected 
too much and that the true mortality 
percentages lie somewhere between the 
corrected and uncorrected values. The 
fact that the agreement between the 
averages of the corrected and the un¬ 
corrected percentages is reasonably 
good, and that the relative virulence of 
the fungi seems approximately the same 
regardless of the method of averaging, 
seems to indicate that a direct com¬ 
parison of the different experiments 
introduced no serious errors. An aver¬ 
age of the uncorrected and corrected 
averages would probably be a fair index 
to the relative virulence of the fungi 
tested in these experiments. 
Table II. — Damping-off of coniferous stems caused by Pythium debaryanum and 
other Phycomycetes 
Tested fungi 
Number of 
lines tested 
! 
i Number of seedlings 
inoculated 
Per cent of seedlings 
killed 
On Pinus resinosa 
On Pinus banksiana 
1 OnPiceaenglmanni 
I 
Pinus resi¬ 
nosa 
Pinus banksiana 
Picea engelmanni 
Pinus resi¬ 
nosa 
Pinus banksiana 
Picea engelmanni 
Average 
Agar inoc¬ 
ulum 
Rice inoc¬ 
ulum 
Agar inoc¬ 
ulum 
Rice inoc¬ 
ulum 
Pythium debaryanum, all lines_ 
!'Ti 
9 
S 9 
330 
460 
160 
230 
57 
90 
98 
97 85 
Line No. 131 and its reisolations_ _ 
4 
3 
3 
80 
120 
50 
80 
62 
91 
100 
96 87 
Linesisolatedfromconiferoushosts 
6 
4 ! 
4 
120 
180 
70 
100 
55 
86 
100 
98 85 
Lines isolated from other hosts._ _ j 
5 
5 
1 5 
210 
280 
90 
130 
58 
92 
97 
94 85 
Rheosporangium aphanidermatus. J 
2 
2 
40 
40 
40 
58 
83 
95 
_79 
Phytophthora sp.. J 
3 
3 
70 
70 
40 
14 
34 
25 
.24 
Phytophthora cactorum_i 
1 
20 
20 
0 
15 
8 
Pythium artotrogus.. 
3 
lj 
- 
60 
60 
20 
0 
2 
0 
-.,J 1- 
Unidentified Phycornycete_ 
1 
20 
20 
0 
0 
0 
Mucor racemosus.____j 
1 
10 
10 
0 
0 
0 
Controls__ 1 
1 
975 
i 
1, 225 
450 
300 
0.1 
2.5 
0.7 
0.3 1- 
' t 
OJ 
O., 
o 
83 
83 
84 
83 
72 
27 
13 
1 
0 
D 
3 Obviously the percentages could not be corrected in exactly the same way however. For instance, if 
in all the experiments the standard lines caused an average loss of 60 per cent, but in experiment B only 40 
per cent, a proportional correction of each mortality percentage in experiment B would raise a fungus 
which had caused 80 per cent damping-off to 120 per cent. Even in cases where there is no such obvious 
absurdity resulting from correction, the inevitable cramping of the percentage scale near 0 and 100 would 
necessitate some other method of applying corrections. The percentages on which the averages in the 
last columns of Tables II, IV, V, and VI were based, were corrected by the following method suggested 
by Dr. Sewall Wright. 
It is evident that in experiment B of the above hypothetical case a mortality of 40 per cent caused by a 
particular fungus should be corrected to 60 per cent. In the normal frequency curve between the 40th and 
60th percentile the abscissal distance on the base line is 0.5068<r. A 20 per cent loss caused by another fungus; 
in the same experiment was corrected as follows: 
By reference to a normal probability integral table (1, Table IV) it was found that an abscissal distance 
of 0.5068<r on the base line to the right of the 20th percentile had over it 17 per cent of the area of the normal 
probability curve. The 20 per cent mortality was therefore raised to 37 per cent. The same procedure 
raised 10 per cent to 20 per cent, 50 per cent to 69 per cenL and 90 per cent to 96 per cent. The basis of the 
method is the assumption that the conditions which affect a given lot of seedlings infected with a given 
f ungus have a certain standard deviation which is the same for all lots. 
