IN THE TROPICS. 
813 
pod produced two fully green seeds with smooth surfaces. 
One of these seeds germinated and produced a plant which 
bore only eight seeds, all of which were green ; five of these 
were smooth, one wrinkled, and two small and irregular. In 
F 3 only one of the smooth seeds germinated, and the plant 
yielded 50 seeds, all nearly smooth and green. This case is 
plainly exceptional, and agrees with similar exceptions seen 
by Tschermak (58), whose explanation—an unexpected 
false hybridism in respect of yellow—may hold in this case 
as well ; but the evidence is meagre. 
Of the 25 yellow seeds of F 15 20 produced plants bearing 
205 seeds :—sm. y. 109, sm. g. 43, wr. y. 40, wr. g. 13, or in 
the proportion 9 : 3’55 : 3*3 : 1*07. 
The reciprocal cross yielded in F 15 20 seeds all smooth and 
yellow. 14 plants in F 2 produced :—sm. y. 110, sm. g. 44, wr* 
y. 28, wr. g. 14. 
The offspring of yellow seeds in all four cases (Ill.a and 
Ill.b.) were therefore as follows in F 2 :—sm. y. 465, sm. g. 182, 
wr. y. 155, wr. g. 53 ; the proportion being 9 : 3'52 : 3 : 1*02. 
The excess of smooth green seeds over theoretical expec¬ 
tation is distinct in every case, but considering the small 
total number of plants examined this is possibly fortuitous. 
In counting the seeds for the above estimations, green and 
yellow seeds were almost always clearly distinguishable, the 
colour of the cotyledons being recognizable through the 
semi-transparent testa. There was, however, sometimes a 
slight doubt in the discrimination of smooth and wrinkled 
seeds, both of which categories show considerable variation in 
external form.* But a further generation showed that the 
amount of actual error was not very great. Thus 4 seeds out 
of 174 appeared to have been wrongly sampled in this respect. 
Taking all four cases together, and writing A= smooth, 
a = wrinkled, B = yellow, and b = green, the result shown in 
the following table was obtained in F 3 , self-fertilization 
being allowed to take place. 
* Gregory (39) has shown that in snch cases there is still a very clear 
distinction between the starch grains of the two types. 
