256 A Biometrical Studtj of Conjugation in Paramecium 
the Paramecia in each local culture unit. Then, even though the pairing were 
quite at random in each locality, yet if the records for several such localities 
were mixed a spurious homogamic correlation would arise. Now the samples 
used in this work were taken in just such a way as would make most pronounced 
any spurious correlation due to local differentiation resulting from place or time 
factors. Small samples — a drop or two of culture fluid — were taken from different 
parts of the culture at intervals of time. But it is clear that if the observed 
homogamic correlation were spurious and due to this " local differentiation " factor, 
we ought to get sensibly as high values if we consider as a pair the two individuals 
lying nearest in the field of view to each pair of conjugants measured. Such 
pairs will have come from the same environment and have been killed at the 
same instant as the actually conjugated pairs. If such random pairs show no 
homogamic correlation, then it seems to me that it will have been demonstrated 
that the observed homogamy is not a spurious correlation due to local differentia- 
tion in the culture. 
The results of these experimental random pairings are shown in Table XVII. 
TABLE XVII. 
Length-Length Correlation in Random Pairs of Non-Gonjugant Paramecia. 
Series 
Characteis 
r 
No. of 
pairs 
A 
Length (Non-Conjugants) ... 
Length (Non-Conjugants) 
- •0894-t- 
•0462 
105 
>> 
•1204-1- 
•0459 
105 
>} 
)> 
(nearest individuals to Conjugants) 
» 
•0908 ± 
•0462 
105 
)5 
(Non-Conjugants)... 
)J 
(Conjugants 
- •0167-t- 
•0465 
210 
» 
)) 
?) 
)) 
•0407 + 
•0464 
210 
c 
?> 
)) 
(Non-Conjugants) 
- -0575 + 
•0473 
101 
>> 
)) 
(nearest individuals to Conjugants) 
)) 
)) 
•1190-*- 
•0468 
101 
)J 
(Non-Conjugants)... 
?J 
(Conjugants) 
- •0256-1- 
•0474 
202 
Here again we reach the same result as before, that random pairings, however 
made, among the individuals living in the same culture at the time of a conjugation 
epidemic give rise to correlations between the members of the pair sensibly equal 
to zero. In Table XVII., four out of the eight coefficients are plus and four minus, 
and none is significant in comparison with its probable error. These are just the 
results to be expected if the true value of the coefficient of correlation for random 
pairing is zero and we take a number of random samples to determine it. We 
must conclude, I think, that the high coefiicieut of assortative pairing observed in 
the actually conjugated pairs cannot be accounted for by any general assumption 
that there is so little variation in the conjugaiit population that random pairing 
will produce high correlation. Random pairing in the identical populations in 
which we find high homogamic coefficients actually gives rise to zero coefficients. 
Nor can it be said that the observed homogamy is due to the fact that small 
samples were taken at different times from different parts of the culture, and 
