248 Gates. 



derived from certain individuals need not be a matter of great surprise. 

 In any case, the ratios from cults. 53 and 54, which are widely different, 

 being respectively about 3 : i and 3 : 6, should be alike according to 

 Mendelian theory (i). 



Cult. 77 furnished further evidence that each individual displays 

 a certain degree of prepotency, i. e., is capable of producing dwarfs 

 in a certain ratio. This culture is derived from a cross between the 

 two individuals (Nos. XI. 9 and VII. 5) which were selfed in Cults. 53 

 and 66 respectively. Now the ratio T -.t iox the two Cultures 60 and 66, 

 together with the ratio 38 : 5 of the grandiflora grandparent (all of 

 which should be alike on my hypothesis), is. 159 : 24. If this be added 

 to the ratio 60 : 22 of culture 53 (the other parent) we get a total of 

 219 : 46, giving the ratio 4.76 : i, which is not very far from the ob- 

 served ratio 5.6:1. Though the numbers are small, we can at least 

 see that the results are in accord with my hypothesis, whüe they are 

 apparently impossible on the Mendelian basis. Moreover it has never 

 been claimed, so far as I am aware, that dwarfing as a unit-character 

 ever depends upon more "factors" than one. The wide variation in 

 prepotency with regard to R, is also found in the case of T. The two 

 series of results are thus consistent with and help to explain each other. 



Regarding the external appearence of the dwarfs, the following 

 facts are important, (i) Their characters were the same in all the 

 cultures, all belonging to one very characteristic type, though they 

 varied greatly in width of leaf, even on the same plant. They agreed 

 entirely in habit, reaching a maximum height of about 80 cm, while 

 the tails reached over 120 cm. The dwarfs were very peculiar in 



(') The results obtained be de Vries (1908) in his nanellu crosses may perhaps 

 be explained by assuming different degrees of prepotency to exist in different classes 

 of germ cells. Thus he found that rruricala x nanella gives the twin types laeia 

 and veliitina in F,, and that both these tj-pes are heterozygous as regards their egg 

 cells but homozygous in their pollen, the velutina pollen when crossed back with 

 nanella giving only dwarfs. Laeta pollen, when crossed back with nanella, also gave 

 only dwarfs, but laeta when self-pollinated gave only tails, though velutina self- 

 pollinated produced about 60% dwarfs and 30% tails. This behaviour of laeta is 

 paradoxical if we consider it in terms of the distribution of fixed and unchangeable 

 units T and /, for with this sjTnbolism, in crosses with nanella t is dominant over T, 

 while in selfing laeta T is dominant over /. But if we assume that the difference 

 between dwarfing and tallness has a quantitative basis, and that laeta pollen is 

 germinally intermediate between these two conditions, then the paradox disappears. 

 For in nanella x laeta pollen the algebraic sum = dwarfs, because insufficient to 

 produce tails, while in laeta dwarf egg cells x laeta pollen the algebraic sum is on 

 the side of the tails. 



