264 A Biometrical StKfhj of Conjugation in Paramecium 
closel}' as the lengths. But as a matter of fact, as we have seen, thei'e is no 
approach to equality between the direct homogamic correlations for length and 
breadth. Those for breadth are always lower. 
In order to make, so far as possible, a decisive test of whether the observed 
homogamic correlation is due to an equalization in size of the two members of 
the pair during the pi-ocess of conjugation itself or is, on the other hand, a result 
of some process such that it exists at the moment of pairing, an investigation based 
on the following considerations was carried out. 
If the observed homogamic correlation is due to any process of equalization in 
size during the act of conjugation itself, then clearly we should expect to find 
a higher degree of correlation between the members of pairs in late stag^es where 
the individuals would shortly separate than in early stages where the union had 
just been accomplished. To test this question, 70 pairs of conjugants in very early 
stages and 77 pairs in late stages were measured. The basis of the separation 
into "early" and "late" stages was the nuclear condition. The limits chosen 
as well as other data regarding these two series — designated respectively as 
" Series F, Early " and " Series F, Late " — have been given in detail earlier in 
the paper (p. 221) and need not be repeated here. The important thing to keep 
in mind is that, taken as a whole, the individuals of the F^ series were recently 
joined pairs, and those of the Fx, series were on the whole nearly at the end of 
the process and would soou have separated. In Table XX. the constants of these 
two series are compared*. The differences are taken as positive when the 
" late " series is in excess of the " early," and negative when the opposite con- 
dition obtains. 
We see at once that the " late " series does not significantly differ from the 
" early " in anything but the means, and there the difference is only barely 
significant. Both series are equally variable, and show equally high homogamic 
correlation. Instead of the " late " series giving the higher values for this corre- 
lation, as we should expect on the equalization hypothesis, the " early " series 
actually has slightly the higher coefficients ' The differences are of course not 
significant in comparison with the probable errors. In fact no closer equality 
between two random samples of the same population could be desired. The 
results show that in these samples at least there was no significant change in 
the size relations of the body associated with the nuclear changes of the con- 
jugation process during that portion of it studied. If any ecjualization in size 
between the two individuals occurred, it must have taken place very shortly after 
the union, and have been a very brief process. Both of these things seem im- 
probable on biological grounds. 
In order to get further evidence on the matter some direct observations on the 
living organism were made, though on account of lack of sufficiently abundant 
* It should be stated that the fundamental tables for the F series are grouped in units of the ocular 
micrometer. The constants deduced were multiplied by 8'G, the value of a micrometer unit in microns, 
to obtain the means and standard deviations tabled. 
