T hiselton-Dyer . — Morphologica l N otes. 439 
exists elsewhere in the genus but is latent in both parents. 
In attempting to explain how this comes about, I cannot 
better the theory of Delage : ‘ II pent arriver qu’un caractere 
vraiment latent revienne au jour. Le croisement jette un 
grand trouble dans revolution de l’ceuf, par ce rapport 
d’innombrables gemmules inattendues ; certains caracteres 
normaux sont aussi contrairies dans leur developpement et 
des caracteres anciens se developpent a leur place ’ (Structure 
du Protoplasma, 580). 
2. K. Bentii ? x K. flammea £ . A number of seedlings 
were raised which at first differed in no appreciable character 
from seedlings of the same age of K. Bentii as represented in 
Plate XXII. All were exactly alike, and exhibited at first no 
trace of hybrid origin. But though raised at the same time 
and subjected to exactly the same treatment as the reverse 
cross described above (which may be called K. kewensis')^ they 
at once showed a marked constitutional difference in the 
extreme slowness of their growth. When K. kewensis was 
three feet high, the plants of the reverse cross had only 
attained six inches. It is, however, interesting to note that 
after cultivation for two years and a half they have begun to 
develop the same pinnatisect leaves which are so characteristic 
a feature in K . kewensis . As none of the plants have yet 
flowered, what will happen then can only be a matter of 
conjecture. It is probable, however, that they will resemble 
those of K. kezvensis , for as Darwin observes : ‘ Hybrids raised 
from reciprocal crosses . . . rarely differ in external characters ’ 
(Origin, 6th ed., 244). 
It is clear that in the reciprocal crosses the influence of 
K. Bentii has been prepotent. This is in agreement with the 
facts cited from Gartner by Darwin in regard to Nicotiana 
(Variation of Animals and Plants, ii. 67). 
But K. kewensis exhibits in the most striking way a char- 
acter in the foliage which cannot be attributed to either 
parent. If I am right in attributing this to reversion, it 
presents a striking exception to the principle laid down by 
Gartner and accepted by Darwin : ‘ Reversions rarely occur 
