254 JOUENAL OF THE ROYAL HORTICULTUEAL SOCIETY. 
43 per cent, of the second group, about forty flowering plants being 
concerned. In the third generation the progeny of the cruciata flowered 
examples proved constant ; in the fourth, however, a separation again 
took place. , 
Such separations in the later generations have the value of atavism, 
and may probably throw some light on this enigmatical phenomenon. 
A second example I give from a crossing of CEnothera ruhrhiervis $ 
and Oil. nanella^ , which I effected in 1893. The first generation yielded 
only two types, that of the mother, and in lesser number (about 8 per cent.) 
that of the common progenitor CE. Lamarckiana ; nanella forms and 
rubrinervis-nanella combinations were absent. Of both types, at the 
flowering time, several were fertilised with their own pollen. In the 
next generation it then appeared that in both types there were some 
constant and some separating crosses. The separating ruhrinervis plants 
yielded thereby in a progeny of about 300 individuals 18 per cent, 
dwarfs, which at the same time displayed the characters of Oil. ruhrinervis, 
i.e. ruhrinervis -nanella. So far this combination of forms has remained 
constant in subsequent sowings. Nanella plants, which were not at the 
same time ruhrinervis, were not absent in these trials, but were rare. 
If no pure types be crossed, but only types already obtained by crossing, 
the non-uniformity in the first generation becomes naturally increased. I 
obtained by exchange an CE. cruciata, Nutt. x , which agrees thoroughly in 
all characters with the species so named, and, with the exception of the 
cruciata character, comes true from seed. With the pollen of this I 
fertilised in 1898 several plants of CE. Lamarckiana. From the crossed 
seed I obtained a culture of about 300 plants of four types. One was 
quite common, and formed about 94 per cent, of the individuals. It 
strongly resembled CE. muricata, but with looser spikes and with flowers 
the size of CE. biennis. The cruciata character was absent throughout. 
Some individuals were self-fertilised : two of them turned out to be per- 
fectly constant in their progeny of over 100 individuals ; a third did so 
likewise, with the exception that the cruciata character reappeared in 
seven out of ninety- one plants. 
The other three types of the first generation more or less resembled 
CE. biennis. One was barren, another, coloured red-brown like CE. cruciata, 
Nutt., was constant in a progeny of about 130 seedlings ; the third, how- 
ever, separated. This yielded only a few seeds when self-fertilised, from 
which only fifteen plants resulted ; these were, some of them, similar to 
CE. biennis, but red-brown, some similar to the mother, and some formed 
other combinations of the grandparental characters. So far the cruciata 
flowers have failed entirely to appear in the offspring of the two last cited 
types. 
From these and many further experiments it results that crossings 
of unequal heritage in the genus CEnothera present a greater multi- 
formity than has been described with respect to other groups. The first 
generation is sometimes uniform, but often of two or more types. The 
crosses of this generation when self- fertilised are as a rule constant 
through their seed ; sometimes, however, there arise, among the constant 
ones, also some which separate again. 
These last separate then, not into equal, but into unequal parts. 
