Feb. 2,1924 
Tissue Fluids in Cotton 
3 i 7 
two horizontal rows of constants, for example, show the correlations 
between the first collection of Meade and the second collection of Pima 
and the correlation between the first collection of Pima and the second 
collection of Meade for the same triplet. 
To bring out clearly the relationships shown by these correlations 
it is necessary to determine the differences between the two coefficients 
of a pair. In determining and interpreting these differences we must 
bear in mind the fact that the mathematical coefficient of correlation 
may be either positive or negative in sign. Positive coefficients indicate 
that in the long run the two variables tend to deviate from the averages 
of their respective means in the same direction. By higher coefficient, 
as used in the comparison of the two coefficients of a pair in the present 
connection, we understand that which is numerically larger when both 
have the same sign, or the one which most nearly approached +1 when 
the sign of one or both of the coefficients is negative. 
The differences in Table XXII are so taken that a positive sign indi¬ 
cates that the correlation is higher when (a) the Egyptian determination 
is made in the second series and the Upland determinations made in the 
first series, and (6) when the Egyptian determination is made in the 
second series and the hybrid determination in the first series, and (c) 
when the hybrid determination is in the second series and the Upland 
determination in the first series. 
Limiting our attention for the moment to the cross correlations be¬ 
tween the two parent types (i. e., omitting the correlations in which 
one constant is based upon a hybrid individual), we note that the co¬ 
efficients are preponderantly positive in sign, thus showing that when 
a first determination on one variety is compared with a second determina¬ 
tion made at a later date on a sample of tissue of another variety drawn 
from the same part of the plot there will be a measurable similarity 
between them. Individually the constants are low and can not in general 
be considered significant in comparison with their probable errors. 
The correlations in which one of the two variables is a constant based 
on a hybrid plant are very irregular and can not in general be considered 
significant in comparison with their probable errors. 
A comparison between the correlations for the two parent types 
shows that in 8 of the 12 comparisons the correlations are somewhat 
higher when the Upland culture furnished the first sample and the 
Egyptian culture furnished the second sample than when the reverse is 
true. 
We have at present no explanation to offer for this difference in the 
behavior of the two types of plants. The results are set forth merely 
as matters of fact ascertained in the determination of correlations to be 
used primarily in the determination of probable errors. (See p. 289.) 
The explanation of the facts must await further research. 
The correlations between the differences between the biological forms 
compared in the first series of determinations and the differences between 
the same forms as compared in the second series of determinations are 
given in Table XXIII. These are presented because they have been 
used in the determination of the probable errors of the differences be¬ 
tween the differences. 
The second problem, that of the relative magnitude of the coefficients 
for the four physicochemical measurements, will be considered when 
the problem of the magnitudes of the correlations for the same duplet 
or triplet is discussed. (See p. 319.) 
