Anderson and Schafer. — Species Hybrids in Aquilegia. 643 



able similarity of the hybrids. Theoretically we might have expected half- 

 sister hybrids to show half the divergence of their unlike parents. That is 

 to say, if we call the three species A, B, and C, and the Fj hybrid between 



Aiiuilfgts eed/i-frj^a 



Ai;ut/f<!io A c/irysanfAa 



Aijuifegia fulgaris 'carrtpae^a ' 



Fig. 2. Dissections of A. ecakarata, A. ' chtysantha'' (garden form), and A. vulgaris ' com- 

 pacia ' (garden form), with the Fj hybrids between them. The hybrids are all very much alike, and 

 resemble vulgaris much more than they do either of the other two species involved. 



A and B is represented as AB, and that between A and C as AC, and so on, 

 we might have expected the following relations to hold : 



i(A-B) = AC-BC. 

 \ (A-C) = AB-BC. 

 \ (B-C) = AB-AC. 



A mere glance at Fig. a is enough to show that this is clearly not the 

 case. The three hybrids are very much like each other, in spite of the great 

 differences between their parents. If we test the matter objectively, our 

 opinion is adequately confirmed. 



Not only are these three hybrids much more alike than we might reason- 

 ably have expected, but they are more like A. vulgaris than they are like 

 any of the other species involved. This is equally true of the other crosses 

 studied. An attempt was made to measure the resemblances more exactly, 



