Pollen Mother-cells of Lilium canadense. 251 
under case (3), but that the chromosomes are the units whose redistribution 
takes place according to the Mendelian law. This possibility has been 
discussed in connexion with Sutton’s (’ 03 ) hypothesis. It is not out of 
harmony with the facts already referred to of the correlation of individual 
characters. Thus Mendel (’ 65 ) found that in Pisum a grey, grey-brown, or 
leather-brown colour of the seed-coat always occurs in connexion with 
violet-red blossoms and reddish spots in the leaf-axils. On the other 
hand, this hypothesis, at least in the form in which Sutton stated it, seems 
to be inapplicable to the cases of constant hybrids ; and it is negatived 
by the cytological facts in the lily, and in other cases in which a double 
longitudinal splitting occurs. The possibility remains, as I have suggested, 
that radically different methods of chromosome reduction prevail in different 
groups of organisms, although before the fact of such variation can be finally 
accepted it must be confirmed by much better evidence than is now available. 
5. It is possible that portions of the idioplasms which do not fuse, or 
the idioplasms as wholes, may interact, perhaps chemically, upon each 
other while in intimate contact in the fusion thread, so that when they are 
afterward separated any particular portion has not the same hereditary 
value that it had before. Such interaction might conceivably result in the 
parental idioplasms, though remaining separate, coming to resemble each 
other more or less closely ; and in this case the results of their separation 
in the germ-cells, so far as concerns the hereditary endowment of the 
offspring, might be indistinguishable from the results of a complete fusion 
and subsequent equation division (hypothesis 1). The reactions that may 
be imagined to occur in this way belong in the same category as those to 
which I have referred as not impossible in the somatic nuclei. Such 
reactions between the substances of the parental idioplasms would in any 
case produce results extremely difficult to trace in the hereditary characters 
of the next generation ; and in the present state of our knowledge of the 
facts of heredity it is probably quite impossible to test the question of 
their occurrence. 
6. It is also possible that, after their contact in the fusion thread, the 
parental idioplasms may be completely separated by the longitudinal 
splitting, so that each germ-cell receives the pure idioplasm of one or the 
other parent. This would result in something similar to Cannon’s con- 
ception of the ‘ purity of the germ-cells.’ Such a complete separation 
might be a regular occurrence, or it might occur as an individual case under 
Mendel’s law, since one of the possible recombinations of idioplasmic units 
would be identical with the previously-existing combination. This also is 
a possibility which, so far as I can discover, has no basis in the observed 
facts of heredity. It is strongly suggested, as has been said, by the striking 
resemblance between the appearance of the thread when splitting and its 
appearance at the previous stage of fusion. But it would be extremely 
T 
