DIVISION OF CENTROMERE 539 



anaphase. Moreover, the chromatids, which are free elsewhere, are 

 always united at the centromere, and therefore presumably by the 

 centromere, until the second metaphase. This is clear in Drosophila 

 even at the first metaphase (Dobzhansky, 1934; D., 1934). The 

 centromere, therefore, divides at metaphase in mitosis but (owing to 

 the precocity of the first metaphase) not until second metaphase at 

 meiosis (D., 1932). The best test of this view is functional. If we 

 follow its behaviour at mitosis and meiosis (cf. Table 76) we see 

 that its orientation on the spindle depends, either on its relationship 

 with another centromere, giving co-orientation, or on an internal 

 change which the centromeres of univalents undergo to permit their 

 auto-orientation between metaphase and anaphase. Moreover, the 

 mitotic chromosomes at metaphase are no longer capable of co- 

 orientation. When they are united by pseudo-chiasmata they con- 

 tinue to show auto-orientation instead of co-orientation (White, 



1935). 



Since co-orientation depends on non-division and pairing, and 

 auto-orientation always goes with division, the conclusion follows, 

 as we have seen, that the internal change permitting orientation is a 

 preparation for division which is functionally equivalent to it. This 

 pre-division may be described as polarisation. The centromere pre- 

 sumably becomes hour-glass shaped. The fact that orientation does 

 not occur until anaphase in univalents at meiosis shows that 

 polarisation itself does not occur before. Nor can it be supposed 

 that the separation of centromeres is determined by the association 

 of the chromatids in pairs, for, as we have seen in the diplochromo- 

 somes, produced by arrest of mitosis, the chromatids can be asso- 

 ciated in fours by the still undivided centromeres. On the other 

 hand, the normally inert centromere of a daughter univalent may 

 behave in an exceptional and remarkable way : it may divide 

 transversely to give two fragments at second anaphase (Nishiyama, 

 1933 b ; Philp, unpub., on Avena). This is significant in indicating 

 different internal organisations and different methods of reproduc- 

 tion in the centromere and in the rest of the chromosome thread. 

 The centromere, therefore, is not a gene, sensu strict 0. 



These observations go to show that the centromeres resemble the 

 centrosomes in their permanence, in their dimensions (so far as these 



