Using a derivative of the plasmid pBR322 we have tested the dimer catastrophe hypothesis of plasmid instability. Most of the theory was confirmed by our observations, but our data suggest that some of the quantitative aspects need modification. In a recF strain of Escherichia coli we estimated the difference in loss rate between the plasmid in the monomeric and the dimeric state to be a factor of 13-14 and the difference in the loss rate between the plasmid in the monomeric and the trimeric state to be a factor of 14-50. We were able to confirm that plasmid oligomers were heterogeneously distributed within a rec+ population, but we were unable to detect any pronounced difference in the level of growth inhibition exerted by the plasmid when in the monomeric, dimeric, or trimeric state. This leaves open the question as to whether runaway plasmid multimerization was prevented (i) by a small correlation between the inhibition of growth and the 'multimeric status' of the plasmid, (ii) by intramolecular homologous recombination, or (iii) whether the process of runaway multimerization is too slow to be recognized within the duration of the experiments, i.e. 200 generations of growth.