CHIASMATA OF MULTIVALENTS 123 



(Dark, 1932). They also occur in Datura (Belling, 1927) and 

 Triticum (Kihara, 1929 et al.). 



Thus each end of a chromosome in a triploid or in a tetraploid 

 may be associated with all of the ends (two in the triploid and three 

 in the tetraploid) that are homologous with it, or with some of them 

 or with none at all. In other words, the kinds of terminal association 

 at diakinesis correspond exactly with the kinds of interstitial 

 association (by demonstrable chiasmata) observed earlier. Ten 

 kinds of quadrivalents can be expected in this way, and they have 

 all been found in Datura (Belling, 1927 h) or Primula sinensis (D., 

 1 93 1 a). Clearly types where the chiasma is formed in each arm 

 should be commoner, where the centromeres are approximately 

 median, than those where two are formed in one arm and none in 

 the other. It follows, therefore, that the simple ring or chain 

 quadrivalents are the commonest types, in forms with median 

 centromeres [Primula, Datura, Campanula). 



In tetraploid Primula not only simple bivalents but very 

 occasionally univalents (unpaired chromosomes) are found. Since 

 the latter do not occur in the corresponding diploid it must be 

 supposed that the exchanges of partner at pachytene are hindered 

 by' the presence of intervening chromosomes to a greater extent 

 than in the diploid (cf. interlocking), and that pachytene association 

 is therefore often incomplete. The time-limit comes into effect. 



(ii) Metaphase : the Limits of Chromosome Association. The 

 extent to which corresponding chromosomes associate at metaphase 

 has been described in a number of polyploids in which these 

 chromosomes are supposed to be identical or nearly identical. In 

 several triploids and tetraploids the chromosomes have been said 

 to associate /w%, i.e., in x trivalents or quadrivalents [e.g., Canna, 

 Belling, 1921 ; Datura, Belling, 1927 ; Primula Sieboldii, Ono, 

 1927 ; Hemerocallis, Sinoto, 1929 ; Hyacinthus, Belling, 1928). 

 Recent work has shown that, while full association does occur in 

 most such forms in a proportion of nuclei, in a proportion also it 

 fails. Trivalents are replaced by bivalents and univalents in the 

 triploids and by pairs of bivalents in the tetraploids. Associa- 

 tion is therefore both incomplete and variable, as Table 15 

 shows. 



