I93I The secondary association of chromosomes 373 



pairs, the probabilities are that competition in chiasma-formation will 

 exaggerate this difference and ultimately lead to bivalent in place of 

 quadrivalent formation. Thus in the diploid Primula kewensis, verti- 

 cillata regularly pair with floribunda chromosomes, but in the tetraploid 

 this rarely occurs. In short, we shall find that although correspond- 

 ing chromosome types in the allopolyploid do not normally pair, they 

 will have a general affinity for one another resulting in secondary 

 association. 



In the auto-polyploid, secondary association may be expected be- 

 tween chromosomes which have failed in multivalent associations. 



The consequence of the high survival value of chromosome differ- 

 entiation in the allopolyploid is a progressive evolution in the direction 

 of functional diploidy. This will be true not only for orthoploid species 

 but for forms (like Pyrus) in which chromosome types are represented 

 in varying numbers. 



c) Secondary Association as a criterion of homology 



The ultimate consequence of the progressive differentiation refer- 

 red to above is the establishment of species with a new basic number. 

 If the affinity of chromosomes is directly proportional to the degree of 

 their identity, the secondary association should be found in varying 

 degree, and generally we might expect this degree to be correlated 

 with the age of the species. 



As secondary association is only observed in plants with small 

 chromosomes it is usually impossible to compare even the major differ- 

 ences between chromosome types, hence in the absence of multivalents, 

 secondary association provides the only criterion of homology available. 

 Thus, ignoring the evidence from the occurrence of multivalents in 

 Dahlia variabilis, the association of the bivalents in groups of 3 and 4 

 is an indication that this species is octoploid and not tetraploid as sup- 

 posed by ISHIKAWA. By the same token D. coronata and D. coccinea 

 are tetraploids and we are able to judge their degree of polyploidy in 

 the absence of any diploid forms. In this respect D. Merckii is partic- 

 ularly interesting insomuch as secondary association completely sub- 

 stantiates the assumed basic number of 8 for the genus Dahlia. More- 

 over in D. coccinea the incidence of the two chromosomes types coupled 

 with secondary association enables one to confirm this conclusion in a 

 striking manner. In the side view illustrated in Fig. 2 g secondary 

 association indicates four chromosome types with two terminal chias- 



