OLD POLYPLOIDS 225 



(Erlanson, 1933), and of Primula ohconica (Philp, unpublished) and 

 Euphorbia granulata (Hagerup, 1932), where there is no evidence of 

 gigantism in the tetraploid forms except in the size of the cells 

 (cf. Allium, Levan, 1932 ; Narcissus, Nagao, 1933). 



3. Finally, the great body of polyploid species show a purely 

 random size relationship with their diploid relatives, being on the 

 average neither larger nor smaller as a whole or in any particular 

 structure. Thus Kihara and Ono (1926) found that the size of pollen 

 grains in Rumex Hydrolapathiim (2n = 200) was no greater than in 

 related diploid and tetraploid species (2n = 20, 40). 



It is therefore necessary to explain why polyploid species have 

 lost the initial size difference they must be supposed to have had, 

 first in their general growth, and later in their detailed structure. 

 Apart from changes in chromosome number, it is found that the 

 progeny of new allopolyploids vary in fertility (e.g., Mgilotricum, 

 Kihara and Katayama, 1931 ; Raphanohrassica, Karpechenko, 

 1927, a and h), and in size [Primula kewensis, Newton and Pellew, 

 1929 ; Nicotiana digluta, R. E. Clausen, 1928). Such changes must 

 be chiefly due to the pairing and crossing-over which must take place 

 occasionally between chromosomes of opposite parents, even with 

 the highest differentiation. Dwarfing, probably from such segrega- 

 tion, has been found in two original tetraploids (Kostoff, 1935 a). 



In old-established polyploids such as Triticum vulgar e and Avena 

 saliva such abnormalities still occur, and when they occur give rise 

 to variation. An occasional quadrivalent is formed owing to the 

 pairing of dissimilar chromosomes from the ultimate diploid ancestors 

 (autosyndesis) as well as pairing between identical chromosomes. 

 When the dissimilar chromosomes pair, segregation of new types 

 (actually resembling related species) results. This " illegitimate " 

 pairing gives, as we have seen, " secondary segregation," and occurs 

 more frequently in varieties newly derived from crossing different 

 races. 



In this way will arise variations on which natural selection can 

 work to change the size of the new form. But such variations will 

 not increase its fertility. Recombination of whole chromosomes or 

 even of any large parts will mean a reduction in the differentiation 

 of opposite sets, tending towards autopolyploidy and hence towards 



R.A. CYTOLOGY. 8 



