At the meta-anaphase transition the centromeres in a genome separate in non-random sequential manner. This sequential separation depends upon the timing of replication of DNA located in the pericentric and centromeric region. Cells in long term cultures as well as some newborn humans carry dicentric chromosomes. The inactive centromeres in these dicentric chromosomes do not show any sequence of separation. Whether or not a dicentric chromosome would segregate equationally depends upon if only one centromere binds to microtubules or both are functional. In man and other higher apes, a 171 base pair long DNA repeat (the alphoid sequence) is present on all centromeres. In mouse, the minor satellite fraction is said to constitute the centromere. These two DNAs also carry a 17 bp long sequence, the CENP-B 'box' to which the CENP-B antigen is bound. Other species-eg, rat, pig, fish, Chinese hamster-exhibit still different sequences at the centromere and do not carry the CENP-B 'box' even though the antigen is ubiquitously present in all species. It is not clear why so many diverse sequences constitute the centromere when all centromeres look alike and perform the same function. I propose that the primary constriction owes its property not necessarily to its DNA composition but to some stereophysical property, eg the curvature and that the region is held together till late metaphase-anaphase due to a specific proteinaceous factor. The mammalian centromeres bind a complex of several proteins dubbed as CENtromere Proteins (CENP's). This complex, however, is not what constitutes the trilamellar kinetochore structure as see under the electron microscope.(ABSTRACT TRUNCATED AT 250 WORDS)