No. 565] CHANGES PRODUCED BY SELECTION 43 



be shown best by an illustration. The 20 leaves of the 

 ' 4 Havana" variety and the first 20 leaves of the " Suma- 

 tra" variety are represented by 10 pairs of factors, of 

 which nine are the same and one different in the two 

 strains. The " Havana" variety is nine leaf factors plus 

 AA, the first 20 leaves of the "Sumatra" variety are nine 

 leaf factors (the same as those in the "Havana") plus 

 BB. The additional leaf factors of the "Sumatra" are 

 CC, BB. EE. With these assumptions, the recombina- 

 tions of a tetra-hybrid will represent our facts fairly 

 accurately. But, as was stated above, it does not seem 

 wise to take this interpretation of the facts too literally. 

 That some such factorial combination will represent our 

 facts superficially there can be no doubt, but in reality if 

 one could grow hundreds of thousands of individuals and 

 follow the behavior of each he would likely find himself 

 constrained to represent his breeding facts by a much 

 more complex system. There would probably be gametic 

 couplings and factorial differences whose main effect 

 would be on some entirely different character or complex 

 of characters, but which would have some slight jurisdic- 

 tion over leaf determination. To become diagrammatical, 

 the unit characters of a house are its cornices, its win- 

 dows, its floors and what not, but a collection of these 

 components is not a house. We may even exchange 

 dormer windows with our neighbor, but we can exchange 

 them only if they fit. Again, we may put on a coat of 

 paint, a color unit, but this color unit affects the appear- 

 ance of many other parts that are just as truly units. 



The essential part of our conception of the origin of 

 this hybrid type is that recombinations of characters 

 quantitative in their nature can be expected and predicted 

 in crosses in exactly the same manner as is done with 

 qualitative characters. On the other hand, it must be 

 borne in mind that here was a hybrid type that appeared 

 to be breeding true to the general characters that we have 

 described, in the F 4 generation. That it was not breed- 

 ing true is clear from the results of the selection experi- 

 ments, yet out of the small number of F 5 and F 6 families 



