Recombinant plasmid pCED3 is structurally unstable in Bacillus subtilis cultures. We have previously shown that stability can be independently increased by changing from a complex medium supporting high growth rates to a chemically-defined medium supporting a lower growth rate and removal of a 4.77-kb EcoRl fragment from pCED3 to give plasmid YSI. Further stabilization was achieved by combining the two approaches. In the present work, we show that the stabilization of the plasrnid-encoded LacZ' phenotype can be explained solely by the effect on the growth rate ratio between cells containing modified and parental plasmids. By using modified stability experiments (where a single cell rather than a suspended colony was used to initiate growth), independent growth rate measurements, and a simple mathematical model, we can describe the kinetics of the loss of the L a d + phenotype in terms of two variables, a and p (where a is the ratio of growth rates between modified and parental cells, and p is the probability of obtaining modified cells from parental cells). Under the conditions tested, the average values of a were 1.52 for cultures growing in complex medium, 1.28 for cultures growing in defined medium, and 1.18 for cultures containing the modified plasmid pYS1 growing in complex medium. The calculated p values ranged between and lo-'' un- der all conditions. Plasmid (pYS137) was used to directly estimate plasmid deletion rates in 8. subtilis and it showed a rate between 5 x lo-* and 1.1 x lo-' deletions/cell/ generation. In contrast to 8. subtilis, there were no detectable differences in growth rates between Escherichia coli strains harboring plasmid pCED3 and plasmid-free cells. These results explain the observed stability of pCED3 in E. coli cultures and indicate that readily detected instability in 8. subtilis cultures can be the result of rare deletion events.