1. A detailed account is given of the behaviour of the chromosomes from diplotene to second anaphase in Dahlia Merckii (2n=36), D. coccinea (2n=32), D. coronata (2n=32) and D. variabilis (2n=64).
2. Two kinds of chromosome association are found in these species:
a) Primary Association, arising from particulate pairing at zygotene and determining segregation at metaphase.
b) Secondary Association, arising at pro-metaphase and due to the general affinity of homologous chromosomes. This does not affect segregation. The chromosomes first associate at pro-metaphase when they are in close proximity to each other.
3. As first described by DARLINGTON, the primary association of chromosomes at metaphase is shown to be solely due to the mainten-ance of chiasmata formed at pachytene. Only those chromosome pairs or multivalents which are materially connected by chiasmata survive the strong repulsion phase of diakinesis. This repulsion results in the radial dispersion of the bivalents to the periphery of the nucleus, coupled with an equal aversion between the pairs and the members of each pair. The repulsion is fully maintained until mid-diaknesis, and then gradually diminishes to pro-metaphase, when the close proximity and general affinity of homologous chromosome results in groups of secondarily associated chromosomes at metaphase. The frequency and size of the group is characteristic for a given species. The importance of the diakinetic Phase in differentiating primary and secondary association is emphasised.
4. The theory of secondary association is discussed and evidence from the literature presented. It is claimed that secondary association is evidence of more remote affinities than can ever be expressed in primary association in a poly-ploid. This is proved by the occurrence of true (primary) association in triploid Pyrus between chromosomes which only show secondary association in the diploid form.
5. The effects of fixation, the correlation between allopolyploidy and secondary association and finally the value of secondary association as a criterion of homology are discussed.