ABSTRACT Investigations of lampbrush chromosomes and the nucleolar organizer have suggested that each gene may be duplicated many times in consecutive linear series within one DNA molecule. This conclusion is in direct conflict with recombination data which indicate, not only that each gene is represented only once per chromatid, but that different genes are contiguous. This paradox is resolved by postulating that the chromosome has the form of a cycloid. Each loop of the cycloid would correspond to a set of copies of a gene forming a chromomere. It is suggested that at meiosis the copies of the gene are detached as a result of intrachromatid crossing-over between the first and last members of the series. The master copy remaining in the chromatid would then be in a position to undergo crossing-over with a homologous chromatid, while the duplicate copies in the detached chromomere would all be included in a single circular DNA molecule. They could subsequently be restored to the chromatid by crossing-over between one of their number and the master copy. This intrachromatid crossing-over would imply that the chromosome can alternate between two states with each set of duplicate genes either detached as a circle or integrated with the DNA axis. Callan’s model for matching slave genes against a master copy so that all acquire identical nucleotide sequences is modified to facilitate coiling and uncoiling of nucleotide chains, by postulating breakage of the matching chains at one end of the gene. Matching of only one chain of the slaves against the master is proposed or, if necessary, subsequent matching of the second slave chain to the first. It is suggested that matching may regularly precede the synthesis of messenger RNA. Investigations of dipteran salivary gland chromosomes and amphibian oocyte nucleoli have established that the chromomere is the unit of replication of the chromosome. On the cycloid model the replicons would be adjacent to one another, and each would comprise a master gene and all the copies. It is suggested that the replicator may correspond to the operator of the master copy of the gene. This hypothesis provides an explanation for several previously unexplained features of crossing-over, including its occurrence at the four-strand stage.