Stable dicentric chromosomes from three mouse cell lines (viz., SEWA Rec4, brain tumor, and L-cells), as well as a human t(9;11) line were analyzed for the sequence in which the two centromeres separate. At prometaphase, as well as in many cells at midmetaphase, the dicentrics express the two centromeres in the form of two primary constrictions. As the cell advances to late metaphase, one of the constrictions loosens the two chromatids so that eventually there is no connection between them. The other centromere stays intact during this period and separates into two units at the metaanaphase junction along with the rest of the genome. The centromere that separates prematurely (out-of-phase) usually is the same in a given dicentric. It is proposed that such a prematurely separating centromere does not function as active element during chromatid migration. Apparently, in dicentrics some sort of control is exerted to eliminate the functioning of one centromere. The nature of such control is not understood at this time. The mouse dicentrics "synthesize" only one kinetochore as definable by antikinetochore antibody studies.