Plasmids containing oriP, the latent origin of replication for Epstein-Barr virus, support efficient replication in selected cell clones when the viral protein EBNA-1 is provided, being lost at a rate of 2 to 4% per cell generation after removal of selection (A. L. Kirchmaier and B. Sugden, J. Virol. 69:1280–1283, 1995; B. Sugden and N. Warren, Mol. Biol. Med. 5:85–94, 1988). We refer to these plasmids as established replicons in that they support efficient DNA synthesis and partitioning each cell cycle. Unexpectedly, we have found that upon introduction of oriPplasmids into a population of EBNA-1-positive cells, oriPplasmids replicate but are lost precipitously from cells during 2 weeks posttransfection (>25% rate of loss per cell generation). Upon investigation of these disparate observations, we have found that only 1 to 10% of cells transfected with an oriP plasmid expressing EBNA-1 and hygromycin phosphotransferase give rise to drug-resistant clones in which the oriP replicon is established. A hereditable alteration in these drug-resistant cell clones, manifested at the genetic or epigenetic level, does not underlie the establishment of oriP, as newly introducedoriP plasmids replicate but are also lost rapidly from these cells. In addition, a genetic alteration in the oriPplasmid is not responsible for establishment, as oriPplasmids isolated from an established cell clone, propagated in Escherichia coli, and reintroduced into EBNA-1-positive cells are likewise established inefficiently. Our findings demonstrate that oriP replicons are not intrinsically stable in EBNA-1-positive cell lines. Rather, the establishment of anoriP replicon is conferred upon the replicon by a stochastic, epigenetic event that occurs infrequently and, therefore, is detected in only a minority of cells.